Introduction
Bone Marrow (BM), a spongy component containing blood vessels inside bones, is often found in the core of large bones like the hipbone and breastbone [ 1 ]. It exists in two forms: red and yellow, as shown in Fig. 1 . Red BM contains hematopoietic stem cells capable of producing red blood cells, white blood cells, and platelets. Yellow BM is predominantly composed of adipose tissue and houses stem cells that are capable of developing into cartilage, fat, or bone cells [ 2 ]. Hematopoietic Stem Cell Transplantation (HSCT), also known as bone marrow transplantation (BMT), is a surgical procedure in which healthy bone marrow stem cells replace injured or destroyed bone marrow [ 3 ]. Stem cells may be sourced from the patient (autologous transplant), a compatible donor (allogeneic transplant), or umbilical cord blood. This method is frequently employed in the treatment of diseases including leukemia, lymphoma, multiple myeloma, and various non-cancerous conditions such as Aplastic anemia and Sickle-cell anemia [ 4 ]. The following conditions may necessitate a BMT [ 5 ]: BMT may be necessary as a component of the treatment protocol for some cancers, including Leukemia, Lymphoma, and multiple Myeloma. Some people may suffer from bone marrow abnormalities such as aplastic anemia, myelodysplastic syndromes (MDS), or hereditary bone marrow diseases. Sickle-cell anemia, thalassemia, and severe combined immunodeficiency (SCID) are genetic illnesses impacting blood cell formation. In some autoimmune illnesses, the immune system assaults and destroys bone marrow, resulting in reduced blood cell production. Some harsh cancer therapies, such as high-dose chemotherapy or radiation therapy, might harm bone marrow. Fig. 1 Anatomical representation of bone marrow, highlighting its structure and key components relevant to BMT
BMT is a complex therapy that has the potential to both save and endanger lives [ 6 ]. The success of BMT largely depends on thorough evaluation, meticulous data collection, and in-depth analysis. Therefore, it is crucial to gather a significant amount of clinical data before initiating therapy. Medical professionals can greatly benefit from the support of artificial intelligence (AI) systems in various aspects of patient care.
The use of AI in BMT is just one example of the rapid expansion of technology in healthcare [ 7 , 8 ]. AI has become increasingly prevalent in complex healthcare systems, demonstrating its strong predictive capabilities in addressing critical issues. This approach shows significant potential for enhancing patient outcomes and progressing medical research in BMT. For instance: Matching Donors and Recipients is one use of AI in BMT [ 9 ]. AI algorithms analyze various parameters, including genetic compatibility, type of human leukocyte antigen (HLA), and other relevant patient data, to improve the donor-recipient matching process. This increases the success rate of allogeneic BMT by reducing rejection and enhancing survival. Predictive analytics : AI systems can analyze large amounts of patient data, allowing them to anticipate the likelihood of problems and the course of a BMT [ 10 ]. These cutting-edge models offer individualized risk evaluations, enabling healthcare practitioners to make more informed clinical decisions. Treatment Optimization : AI technologies can improve BMT treatment protocols by evaluating data from clinical trials and prior transplant cases [ 11 , 12 ]. Medical professionals may now tailor and enhance treatment strategies according to the unique characteristics of each patient and illness profile. Because these algorithms identify patterns and trends in treatment responses and outcomes, they help medical specialists make well-informed decisions. Monitoring systems : Driven by AI, these systems can efficiently follow a patient’s development before, during, and post BMT [ 13 ]. These cutting-edge systems can track patient data in real-time, including clinical observations, test results, and vital signs. They can then quickly spot early warning signs of possible problems, including transplant rejection, infection, or graft-versus-host disease.
Accurate prognosis and classification in BMT are critical for treatment planning, risk assessment, patient counseling, tracking treatment progress, performing research, and allocating resources efficiently. These characteristics greatly improve overall treatment results and the quality of treatment provided to patients undergoing this operation.
Despite the extensive research on the prognosis and classification of BMT treatment outcomes, numerous challenges persist and demand attention. These challenges are listed in Fig. 2 . Patient heterogeneity : BMT comprises a heterogeneous patient population that varies in age, underlying disorders, disease stages, and genetic variables. This variation makes it difficult to develop standardized prognostic and classification models to predict patient treatment results properly. Disease complexity and variability : BMT treats various illnesses, including genetic disorders, lymphomas, and leukemia. Each disease has distinct features, including multiple subtypes and related risk factors. It is challenging to develop prognostic and classification methods that are broadly applicable due to the complexity and variety of these disorders. Limited data and sample size : Because BMT is a very uncommon and specialized operation, there is a shortage of information accessible, especially when it comes to rare diseases or certain patient subgroups. Consequently, the small sample size can hinder the development of robust prognostic models and classification systems. Prolonged observation and delayed results : To assess treatment results in BMT, continued monitoring is necessary due to the delayed development of some problems and relapses. However, acquiring and interpreting long-term data can be difficult since it requires consistent patient participation and resources. Treatment Improvements and evolving procedures : BMT is a rapidly growing discipline. Advancements have been made in conditioning regimens, donor choice, graft-versus-host disease prevention, and supportive care. The introduction of novel medicines and developing procedures has the potential to significantly alter treatment results, rendering existing prognostic and classification models inaccurate or even obsolete. Interaction of numerous factors : Several factors impact the outcome of BMT, including patient characteristics, disease-related factors, donor factors, and post-transplant problems. It cannot be easy to understand and incorporate the complex interaction of these aspects into prognostic models and categorization systems. Statistical and methodological limitations : Appropriate statistical techniques and extensive validation are necessary for reliable prognostic models and classification schemes. However, problems like selection bias, overfitting, and a lack of outside validation might affect how reliable and generalizable the models are. Fig. 2 An overview of the primary challenges encountered in BMT research, including statistical limitations, evolving treatment protocols, and interpretability issues in AI-driven models
This study introduces a notable advancement in pediatric healthcare, specifically regarding the prognosis and classification of childhood BMT through a sophisticated Computer-Aided Diagnosis (CAD) framework. By integrating state-of-the-art AI and machine learning techniques, the study addresses critical gaps in the current diagnostic processes, offering clinicians a more precise, reliable, and interpretable methodology. This research presents the following key contributions: Innovative CAD Framework : Developed a comprehensive CAD framework specifically tailored for the classification of childhood BMT, enhancing diagnostic accuracy and providing a solid foundation for treatment decision-making. Particle Swarm Optimization (PSO) for Feature Selection : Employed PSO to effectively select the most pertinent features from the dataset, significantly improving the efficiency of machine learning models in predicting transplant outcomes. Bayesian Optimization with Adaptive TPE : Implemented the Adaptive Tree of Parzen Estimators (TPE), a Bayesian optimization technique, to fine-tune hyperparameters of various machine learning models, ensuring optimal model performance. Examination of Scaling Techniques : Conducted a thorough investigation into the impact of different data scaling techniques on model performance, providing valuable insights into data preprocessing in medical data analysis. Integration of Explainable AI : Utilized Local Interpretable Model-Agnostic Explanations (LIME) to ensure the transparency and interpretability of the AI models, facilitating better understanding and trust among clinicians in applying AI-driven diagnostics.
This paper is organized as follows: Section 2 provides a comprehensive assessment of the existing research on BMT, identifying potential research gaps. Section 3 examines the approach, detailing the materials, methods, datasets, and performance metrics employed. Section 4 provides the results and discussion derived from comprehensive testing and analysis. Section 5 summarizes the findings of the paper and suggests potential directions for future research.
Literature review
This section contains in-depth information regarding research trends in childhood BMT, particularly those involving ML and DL techniques. A thorough bibliometric analysis is performed to shed light on the limitations of available data and identify prospective areas. Following that, a comprehensive review of the available BMT research is presented.
Bibliometric study
A comprehensive bibliometric analysis of research trends utilizing deep learning (DL) and ML approaches to help predict survival rates of childhood BMT was carried out from 2015 to 2023. This procedure highlights the limitations of existing data and pinpoints potential areas for future study. The choice of 10 years of research was made to offer researchers a complete understanding of how research has changed over these periods and to aid in understanding the components and characteristics of study patterns. This data is beneficial as it elucidates the patterns in scientific publications among organizations, countries, journals, and authors.
Keywords used for making this search on the Scopus database are “Bone Marrow Transplant, “Children”, “machine learning”, “deep learning”, “artificial intelligence”, and “intelligent systems”. Below are some samples of searches performed on the Scopus database:
(“HSCT” OR “HCT” OR “GVHD” OR “acute GVHD” OR “aGVHD” OR “leukemia” OR “lymphoma” OR “autologous HCT” OR “allogeneic HCT” OR “Hematopoietic Cell Transplantation” OR “Bone marrow transplant” OR “Hematopoietic cell transplant” OR “Hematopoietic stem cell transplantation” OR “Graft-versus-host disease”) AND (“Machine Learning” OR “Artificial Intelligence” OR “Deep Learning”) this search result on 2946 documents ((“Bone Marrow Transplant” OR “BMT” OR “Hematopoietic Stem Cell Transplantation”) AND (“Children” OR “Child” OR “Childhood”)) AND (“deep learning” OR “air” OR “Artificial intelligence” OR “Neural networks”) this search result on 54 documents.
The authors considered the final search since it included AI, DL, and ML terms. Finally, research trends on AI, DL, and ML approaches to help and predict survival rates of Childhood BMT were found through bibliometric analysis of 54 published publications. Several research papers and document results from the Scopus database will be discussed in this section, along with an analysis of them. According to the Scopus database, 54 works in the field of approaches to help and predict survival rates of Childhood BMT were published in various journals between 2015 and 2024. Fig. 1 displays the quantity of papers and documents from 2015 to 2024. According to Fig. 3 , 2021 shows the most publications, with ten articles released. 2022 comes in second with nine papers published. Meanwhile, 2024 saw the fewest publications, with only one paper published yearly. Fig. 3 Published documents and papers by year
During the last ten years, at least one full-length research article on helping and predicting survival rates of Childhood BMT-based intelligent systems (AI, ML, DL) has been published in 54 research outputs from 25 countries. Fig. 4 shows a map of the top 10 countries that, over the last ten years, have contributed most to the development of helping and predicting Childhood BMT research survival rates. Approximately 68% of all publications globally were produced, and their respective participation rates determine the total number of papers published in each country: the United States, China, Italy, and Turkey. The US is the global leader, accounting for 16% of all documents globally. China, Italy, and Turkey contributed 11% for each. Fig. 4 Published documents and papers by country
The distribution of publications by kind of document is seen in Fig. 5 . Article papers have the greatest value, accounting for 81.5% of the total, whereas reviews and book chapters constitute 13% and 1.9%, respectively. Fig. 5 Published documents and papers by type
The distribution of documents by subject area is illustrated in Fig. 6 . Medicine and Biochemistry, Genetics, and Molecular Biology account for approximately 78.4% of the subject areas. The field of Medicine is demonstrated to be the most popular discipline for helping and predicting survival rates of Childhood BMT-based intelligent systems (AI, ML, DL) at 67.6%, followed by Biochemistry, Genetics, and Molecular Biology at 10.8%. Also, Computer science, Chemical engineering, and Engineering subject areas have approximately six % among different subject areas. Fig. 6 Published documents and papers by area
Related studies
Artificial intelligence (AI) technology has significantly improved BMT in recent years. An overview of relevant research is given in this section, with emphasis on recent studies and approaches that have aided in the development of AI-enhanced analysis of children’s BMT. An ML and explainable artificial intelligence (XAI) approach is utilized to predict the survival probability of children receiving hematopoietic stem cell transplantation [ 14 ]. Three methods for feature selection have been used: Harris Hawks optimization, salp swarm optimization, and mutual information. The final tailored stacked model achieved optimal performance metrics, including precision of 89%, recall of 88%, and f1-score of 88%. Moreover, XAI techniques such as ELI5, Shapley additive values (SHAP), QLattice, and LIME have been employed to enhance the models’ precision, comprehensibility, and interpretability.
The goal of the proposed method in [ 15 ] was to create and verify an ML method for forecasting stem cell donors’ accessibility. About 10,258 verified typing inquiries were made over the HSCT procedure in the British Bone Marrow Registry (BBMR) between January 1, 2013, and December 31, 2018. They prospectively gathered a data set encompassing these requests. Several ML algorithms such as logistic regression, support vector machines, and boosted decision trees (BDTs) were applied and contrasted. The AUC was used to evaluate these methods. The findings of this paper demonstrated that BDTs were more successful in forecasting the accessibility of BBMR donors. Using the AUC of the test sample of 2052 data points, the model’s overall predictive power was determined to be 0.826.
Ggaliwango Marvin et al. [ 16 ] use SHAP to computationally apply ML and AI methods for the predictable and interpretable identification of significant factors influencing the efficacy or decline of stem cell transplantation in pediatric patients. Additionally, the authors present a clearer method for predicting pediatric hematologic patients’ survival capacity determined by the significance and strength of associations related to survival-determining features before cell transplantation. The accuracy obtained was 82% for the Catboost algorithm, 92% for LigtGBM, and 94% for XBoost.
The Mud Ring Algorithm (MRA) provides a novel feature selection method for survival prediction challenges [ 17 ]. Thirteen distinct real datasets with varying occurrences were utilized for experimentation and testing to compare the proposed approach with other techniques. Subsequently, the bone marrow transplant children’s dataset was used to assess the built model’s ability to classify against alternative methods. Classification results were obtained using contemporary approaches, and a variety of established metrics, visual tools, and diagnostic techniques were employed to compare them to the recommended results. This study has shown that the recommended strategy is comparable and produces better results than alternative approaches. Furthermore, the outcomes demonstrated that the developed model improved the accuracy of predictions for test instances by as much as 82.6%.
A new method for diagnosing leukemia by accurately evaluating multi-omics data using ML and DL techniques is presented in [ 18 ]. This study analyzed ML methods like Random Forest (RF), Naive Bayes (NB), Decision Tree (DT), Gradient Boosting (GB), Logistic Regression (LR), and DL techniques like Recurrent Neural Networks (RNN) and Feedforward Neural Networks (FNN). RNN surpassed 98% accuracy in DL, compared to 97% in ML for GB.
A survey of the use of ML techniques in HCT was carried out in [ 19 ]. Researchers looked at what kinds of data sources were included, what ML algorithms were applied, and what clinical results were quantified. An in-depth examination of the literature was conducted utilizing databases such as IEEE Xplore, Web of Science, PubMed, and Scopus. Hematopoietic cell transplantation (HCT), autologous HCT, allogeneic HCT, ML, and AI were among the search phrases used. Only full-text research published between 2015 and 2020 was considered. Two writers used specified data fields to extract the data. A collection of 242 papers was found using PRISMA standards, of which 27 studies satisfied the inclusion requirements.
An effective classification model using a publicly available dataset is presented in [ 20 ]. The aim was to forecast the survival rates of children undergoing BMT. In this context, some supervised machine learning techniques were examined using an 80–20 train-test split ratio. The Chi-square feature selection method was used to choose only the best 11 of the 59 database features to ensure prediction with the least time and resources. In addition, the grid search cross-validation (GSCV) method of hyperparameter optimization was used to improve the accuracy of predictions. Using the original and the modified datasets, four experiments were carried out using an array of default and optimized hyperparameters. The analysis indicated that the prediction accuracy of the top eleven hyperparameter optimization characteristics was 94.73%.
Artificial Neural Network (ANN)-based BSA, VSA, and GOA methods are employed to enhance the predictive capacity of the neural network [ 21 ]. The proposed models are applied to a UCI database dataset to predict the responses of children with hematologic disorders to bone marrow transplants. AUC, Receiver Operating Characteristic (ROC), and root-mean-square deviation (RMSD) measurements are used to evaluate and contrast the results of various models. It is known that the ANN-BSA approach produces the best feasible outcomes when it comes to the correlation between the data from the input and output layers (i.e., clinical characteristics and survival rates). This study offers concrete evidence of how ML systems greatly aid medical professionals in estimating an individual-based diagnosis.
The analysis of the HCT data set aims to determine the kinds and clinical effects. Transplanting hematopoietic cells is a highly effective therapy for various illnesses. A study in [ 22 ] examines and records the quantity of HCT procedures carried out in the US between 2014 and 2018, in addition to the outcomes of HCT as recorded by the CIMBTR. According to the data, the number of transplants performed in the US has increased. A 13% rise was seen between 2014 and 2018. The examination of transplants is divided into groups according to the US states, donor type, age, and source of stem cells.
The ROI-BMC-DNNet is a unique analytic methodology proposed in [ 23 ]. The authors proposed a pyramid segmentation model utilizing prior pathological data to autonomously identify regions of interest (ROI) that clinicians typically focus on for counting nuclei and obtaining patch images in well-spread cellular areas. The authors present an alignment strategy for multiscale ROI that is straightforward yet efficient in preserving the pyramid model’s semantic and spatial information correctness. The datasets of scanned WSIs from 120 individuals are used for the tests. Results obtained 92.27% retrieval precision for the high-quality patches with our technique.
The retrospective investigation involved individuals diagnosed with aplastic anemia (AA) or MDS via pathological examination of bone marrow samples which was conducted on individuals who underwent pelvic magnetic resonance imaging (MRI) using the iterative decomposition of water and fat with echo asymmetry and least-squares estimation quantitation (IDEAL-IQ) technique from 2016 to 2020 [ 24 ]. Utilizing measurements of fat fraction (FF) in the right ilium and radiometric characteristics derived from both T1-weighted (T1W) and IDEAL-IQ images, three distinct ML algorithms, namely linear discriminant analysis (LDA), LR, and support vector machine (SVM), were employed to differentiate between cases of AA and MDS. This research had 77 individuals. The average age is 47 years, with a gender distribution of 37 males and 40 females, ranging from 20 to 84 years old.
Several ML classifications are used in the dataset to forecast an individual’s survival status following BMT in [ 25 ]. A BMT is a very risky medical procedure since it has the potential to either preserve the patient’s life or treat them. Eleven classification models DT, XG Boost (XGB), RF, GB, ADA Boost (ADA), Cat Boost (CB), and Extra Tree (ET) were used in this work, utilizing two different methods. While the second method uses Principal Component Analysis and a Fuzzy Discernibility Matrix to choose characteristics for testing and training, the first method considers all 23 features for training and testing. The ADA Boost Classification achieved the highest outcomes in the shortest period, as measured by FDM, with accuracy scores of 0.9523, precision scores of 1, recall scores of 0.9259, and ROC-AUC scores of 0.9629.
A cohort of 770 children underwent their initial allogeneic hematopoietic cell transplantation at St. Jude Children’s Research Hospital between 2000 and 2020, as referenced in [ 26 ]. A predictive model is developed by integrating baseline data encompassing disease, donor, and graft attributes with longitudinal assessment data, which includes blood cell counts, serum chemistries, and measurable residual illness, collected within 30 days before or after HCT. This model utilizes penalized logistic regression to identify patients at elevated risk of mortality within one year post-HCT. The authors analyzed the comparative importance of different input factors. They compared the proposed machine learning model utilizing longitudinal measures with a model trained on baseline data.
A revolutionary, fully automated end-to-end DL pipeline is developed in [ 27 ] to predict genetic abnormalities from single-cell images obtained from scans of traditionally stained bone marrow smears currently under evaluation. This workflow facilitated the collection of over 2,000,000 single-cell images from diagnostic samples of 408 AML patients, resulting in a multiterabyte dataset. Convolutional Neural Networks (CNNs) were subsequently trained on these images to predict various genetic variations associated with treatment outcomes. They also created a temporal test cohort data set comprising approximately 444,000 single-cell images from an additional 71 AML patients.
Individuals of 1751 receiving BMT had records with up to 120 measures per transplant surgery (Shariati Hospital). 22.3% of the exceedingly sparse entries may be processed thanks to collaborative filtering approaches [ 28 ]. The effectiveness of numerous classification techniques trained on estimating the survival status was assessed using ten-fold cross-validation. In terms of predicting the survival rates, AUC was attained at 0.69. What’s more, though, was that the best odds of successful surgeries could be identified with a high degree of precision; for example, 92% or 97% of 74 or 31 patients could be identified, respectively.
An AI-assisted molecular evaluation diagnostic assistance system based on BMT, which includes cell identification, classification, and prediction of leukemia types, is proposed in [ 29 ]. A new YOLOX-s model for cell detection to identify cells accurately and produce single-cell images. They considered categorizing cells as a fine-grained challenge. They created a unique architecture dubbed MLFL-Net that uses multi-level characteristics. The leukemia types are also predicted using a dataset comprising 40 healthy individuals (BM transplant donors) and 40 patients with various forms of acute leukemia. Table 1 compares the various state-of-the-art techniques for helping and predicting survival rates of Childhood BMT-based intelligent systems. Table 1 Comparison between the various state-of-the-art techniques for helping and predicting survival rates of Childhood BMT-based intelligent systems Ref Technique Dataset ACC (%) Pros Cons [ 14 ] Custom stacked model, 187 patients 38 features 89 Improved Prediction Accuracy. Limited dataset details Explainable AI Use of explainable AI Complexity of stacked model [ 15 ] Boosted Decision British Bone Marrow 87 High accuracy in predicting donor availability Focused on donor availability, Trees (BDT), Registry (BBMR), Practical for optimizing donor selection not patient outcomes LR, SVM 10,258 requests Large dataset enhances model robustness Predication accuracy needs to be improved. [ 16 ] Explainable computational Pediatric hematologic — Focus on pediatric patients Lack of specific accuracy metrics limits evaluation pathology (SHAP) patient dataset addresses a critical population Small dataset size reduces robustness no details, likely small Explainable AI aids clinical decision-making Limited methodological details [ 17 ] Mud Ring Algorithm 13 real-world datasets 82.6 Novel MRA feature selection improves efficiency MRA may be complex to implement and understand for researchers (MRA) with varying features. Tailored for pediatric HSCT survival prediction Prediction accuracy needs to be improved. [ 18 ] DL (CNN-LSTM, ResNet) 1690 cases. 98 in DL Integrates multi-omics for comprehensive analysis Not directly focused on HSCT outcomes, High computational cost ML (RF, SVM) 97 in ML High potential for leukemia diagnosis aiding HSCT planning Data complexity may limit clinical adoption [ 20 ] Supervised ML, Chi-square 187 samples 94.73 Chi-square feature selection reduces dimensionality Limited Generalizability Hyperparameter Optimization (GSCV) 37 features Hyperparameter optimization improves performance Interpretability Challenges [ 23 ] ROI-BMC 120 individuals 92.27 Automated analysis of bone marrow images Limited to image-based diagnosis DNNet WSIs High potential for diagnostic support in HSCT Requires high-quality imaging infrastructure [ 24 ] MRI-based ML 77 individuals — Novel use of MRI data for bone marrow analysis Limited interpretability, Limited accuracy details AA vs. MSs Aids differential diagnosis relevant to HSCT Specialized MRI data may limit accessibility. [ 25 ] ML classifications 187 transplantations 95.2 Improved Predictive Accuracy, Fuzzy logic-based approaches are 39 features FDM models offer greater interpretability complex to implement and understand [ 26 ] A predictive model that employs penalized 770 kids — Enhanced Prediction Accuracy, Providing real-time predictions. Single-center data limits generalizability logistic regression Single-center data ensures consistency Predication accuracy needs to be improved. [ 27 ] DL 408 — High accuracy in predicting AML genetics Not directly focused on HSCT outcomes (CNN) patients Supports pre-HSCT risk stratification Requires specialized smear data, High computational cost [ 28 ] Data Mining 1750 92 for 74 Early application of ML to HSCT Data Complexity and Dimensionality (DT, SVM) transplantations 97 for 31 Identifies high-survival patients Need to be generalized on more datasets. [ 29 ] DNN 80 89.53 Automated diagnosis from smears Need to be generalized on more datasets. individuals Supports HSCT-related disease identification Predication accuracy needs to be improved.
Table 1
Comparison between the various state-of-the-art techniques for helping and predicting survival rates of Childhood BMT-based intelligent systems
| Ref | Technique | Dataset | ACC (%) | Pros | Cons |
| [ 14 ] | Custom stacked model, | 187 patients 38 features | 89 | Improved Prediction Accuracy. | Limited dataset details |
| Explainable AI | Use of explainable AI | Complexity of stacked model | |||
| [ 15 ] | Boosted Decision | British Bone Marrow | 87 | High accuracy in predicting donor availability | Focused on donor availability, |
| Trees (BDT), | Registry (BBMR), | Practical for optimizing donor selection | not patient outcomes | ||
| LR, SVM | 10,258 requests | Large dataset enhances model robustness | Predication accuracy needs to be improved. | ||
| [ 16 ] | Explainable computational | Pediatric hematologic | — | Focus on pediatric patients | Lack of specific accuracy metrics limits evaluation |
| pathology (SHAP) | patient dataset | addresses a critical population | Small dataset size reduces robustness | ||
| no details, likely small | Explainable AI aids clinical decision-making | Limited methodological details | |||
| [ 17 ] | Mud Ring Algorithm | 13 real-world datasets | 82.6 | Novel MRA feature selection improves efficiency | MRA may be complex to implement and understand for researchers |
| (MRA) | with varying features. | Tailored for pediatric HSCT survival prediction | Prediction accuracy needs to be improved. | ||
| [ 18 ] | DL (CNN-LSTM, ResNet) | 1690 cases. | 98 in DL | Integrates multi-omics for comprehensive analysis | Not directly focused on HSCT outcomes, High computational cost |
| ML (RF, SVM) | 97 in ML | High potential for leukemia diagnosis aiding HSCT planning | Data complexity may limit clinical adoption | ||
| [ 20 ] | Supervised ML, Chi-square | 187 samples | 94.73 | Chi-square feature selection reduces dimensionality | Limited Generalizability |
| Hyperparameter Optimization (GSCV) | 37 features | Hyperparameter optimization improves performance | Interpretability Challenges | ||
| [ 23 ] | ROI-BMC | 120 individuals | 92.27 | Automated analysis of bone marrow images | Limited to image-based diagnosis |
| DNNet | WSIs | High potential for diagnostic support in HSCT | Requires high-quality imaging infrastructure | ||
| [ 24 ] | MRI-based ML | 77 individuals | — | Novel use of MRI data for bone marrow analysis | Limited interpretability, Limited accuracy details |
| AA vs. MSs | Aids differential diagnosis relevant to HSCT | Specialized MRI data may limit accessibility. | |||
| [ 25 ] | ML classifications | 187 transplantations | 95.2 | Improved Predictive Accuracy, | Fuzzy logic-based approaches are |
| 39 features | FDM models offer greater interpretability | complex to implement and understand | |||
| [ 26 ] | A predictive model that employs penalized | 770 kids | — | Enhanced Prediction Accuracy, Providing real-time predictions. | Single-center data limits generalizability |
| logistic regression | Single-center data ensures consistency | Predication accuracy needs to be improved. | |||
| [ 27 ] | DL | 408 | — | High accuracy in predicting AML genetics | Not directly focused on HSCT outcomes |
| (CNN) | patients | Supports pre-HSCT risk stratification | Requires specialized smear data, High computational cost | ||
| [ 28 ] | Data Mining | 1750 | 92 for 74 | Early application of ML to HSCT | Data Complexity and Dimensionality |
| (DT, SVM) | transplantations | 97 for 31 | Identifies high-survival patients | Need to be generalized on more datasets. | |
| [ 29 ] | DNN | 80 | 89.53 | Automated diagnosis from smears | Need to be generalized on more datasets. |
| individuals | Supports HSCT-related disease identification | Predication accuracy needs to be improved. |
Research gaps
The review of AI advancements in Pediatric BMT reveals significant progress, particularly in machine learning, explainable AI (XAI), and optimization algorithms for feature selection and model tuning. Despite these advancements, the literature indicates areas requiring further exploration: CAD Framework Specialization: Current literature demonstrates the application of various AI techniques in pediatric BMT, yet lacks in detailing specialized, integrated CAD frameworks tailored for this specific medical domain. Feature Selection Optimization: The utilization of PSO for feature selection in pediatric BMT datasets is inadequately explored despite mentions of other optimization techniques like Harris Hawks and salp swarm optimization. The novelty is not merely the use of PSO itself, but rather its comparative efficiency, robustness, and adaptability in handling the complex, noisy nature of BMT clinical data. Hyperparameter Tuning Techniques: The adoption of Bayesian optimization, specifically TPE, for enhancing AI model performance in pediatric BMT remains insufficiently documented. Data Scaling Impact: The literature provides insights into ML and DL algorithms, but falls short of thoroughly investigating the effects of different data scaling methods on model performance within the pediatric BMT context. Application of XAI: While XAI techniques such as ELI5, SHAP, QLattice, and LIME are mentioned for improving model interpretability, their effective integration and impact on enhancing clinical decision-making in pediatric BMT require deeper investigation.
Addressing these gaps is essential for advancing AI’s role in pediatric BMT and improving treatment outcomes for affected children.
Methodology
The current study suggests a CAD framework for classifying childhood BMT. The proposed framework is presented in Fig. 7 . Fig. 7 Schematic representation of the proposed CAD framework tailored for the classification of childhood BMT cases; highlighting its modular architecture and integration of AI-driven techniques
Materials
Childhood BMT represents a critical frontier in pediatric medicine, offering a lifeline for children facing various hematological disorders [ 30 ]. To shed light on the intricate dynamics of these transplants, we turn our attention to a comprehensive dataset encapsulating a myriad of factors influencing transplant outcomes. This dataset, aptly named “Bone marrow transplant: children” , serves as the cornerstone for our exploration into pediatric BMT. It can be accessed from: https://archive.ics.uci.edu/dataset/565/bone+marrow+transplant+children [ 31 ].
The dataset contains 187 transplantation cases with 39 features, encompassing categorical, numerical, and Boolean variables. Its columns (i.e., features) capture diverse aspects of each transplant case. These columns encapsulate essential information, ranging from demographic details to immunological compatibility factors. The dataset columns are presented in Table 2 . Table 2 Description of columns in the “Bone marrow transplant: children” dataset Column Demographic Description Recipientgender Gender Male or Female Stemcellsource - Origin of hematopoietic stem cells Donorage Age Age of the donor at the time of apheresis for hematopoietic stem cells Donorage35 Age Whether the donor age is less than 35 IIIV - Occurrence of acute graft versus host disease stage II, III, or IV Gendermatch Gender Compatibility based on the gender of the donor and recipient DonorABO - Blood group of the donor according to the ABO system RecipientABO - Blood group of the recipient according to the ABO system RecipientRh - Presence of the Rh factor on the red blood cells of the recipient ABOmatch - Compatibility of the donor and recipient according to the ABO blood group CMVstatus - Serological compatibility of the donor and recipient for cytomegalovirus DonorCMV - Prior cytomegalovirus infection in the donor before transplantation RecipientCMV - Prior cytomegalovirus infection in the recipient before transplantation Disease - Classification of the disease (ALL, AML, chronic, non-malignant, lymphoma) Riskgroup - Risk level (high or low) Txpostrelapse - Second BMT performed after relapse Diseasegroup - Type of disease (malignant or non-malignant) HLAmatch - Compatibility of donor and recipient antigens as per ALL international BFM SCT 2008 criteria HLAmismatch - Matched or mismatched HLA status Antigen - Number of antigen differences between the donor and recipient Allele - Number of allele differences between the donor and recipient HLAgrI - Nature of the difference between donor and recipient Recipientage Age Age of the recipient at transplantation time Recipientage10 Age Whether the recipient’s age is less than 10 Recipientageint Age Exact age of the recipient Relapse - Reoccurrence of the disease after treatment aGvHDIIIIV - Development of acute graft versus host disease stage III or IV extcGvHD - Extensive chronic graft versus host disease occurrence CD34kgx10d6 - Dose of CD34+ cells per kg of recipient body weight CD3dCD34 - Ratio of CD3+ cells to CD34+ cells CD3dkgx10d8 - Dose of CD3+ cells per kg of recipient body weight Rbodymass - Body mass of the recipient at transplantation time ANCrecovery - Time taken for neutrophil recovery, defined as > 0.5 x 10 9 / L PLTrecovery - Time taken for platelet recovery, defined as > 50000/mm \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$^3$$\end{document} time_to_aGvHD_III_IV - Duration until development of acute graft versus host disease stage III or IV survival_time - Observation period (if alive) or time to event (if dead), in days survival_status - Status of survival at the end of the observation period
Table 2
Description of columns in the “Bone marrow transplant: children” dataset
| Column | Demographic | Description |
| Recipientgender | Gender | Male or Female |
| Stemcellsource | - | Origin of hematopoietic stem cells |
| Donorage | Age | Age of the donor at the time of apheresis for hematopoietic stem cells |
| Donorage35 | Age | Whether the donor age is less than 35 |
| IIIV | - | Occurrence of acute graft versus host disease stage II, III, or IV |
| Gendermatch | Gender | Compatibility based on the gender of the donor and recipient |
| DonorABO | - | Blood group of the donor according to the ABO system |
| RecipientABO | - | Blood group of the recipient according to the ABO system |
| RecipientRh | - | Presence of the Rh factor on the red blood cells of the recipient |
| ABOmatch | - | Compatibility of the donor and recipient according to the ABO blood group |
| CMVstatus | - | Serological compatibility of the donor and recipient for cytomegalovirus |
| DonorCMV | - | Prior cytomegalovirus infection in the donor before transplantation |
| RecipientCMV | - | Prior cytomegalovirus infection in the recipient before transplantation |
| Disease | - | Classification of the disease (ALL, AML, chronic, non-malignant, lymphoma) |
| Riskgroup | - | Risk level (high or low) |
| Txpostrelapse | - | Second BMT performed after relapse |
| Diseasegroup | - | Type of disease (malignant or non-malignant) |
| HLAmatch | - | Compatibility of donor and recipient antigens as per ALL international BFM SCT 2008 criteria |
| HLAmismatch | - | Matched or mismatched HLA status |
| Antigen | - | Number of antigen differences between the donor and recipient |
| Allele | - | Number of allele differences between the donor and recipient |
| HLAgrI | - | Nature of the difference between donor and recipient |
| Recipientage | Age | Age of the recipient at transplantation time |
| Recipientage10 | Age | Whether the recipient’s age is less than 10 |
| Recipientageint | Age | Exact age of the recipient |
| Relapse | - | Reoccurrence of the disease after treatment |
| aGvHDIIIIV | - | Development of acute graft versus host disease stage III or IV |
| extcGvHD | - | Extensive chronic graft versus host disease occurrence |
| CD34kgx10d6 | - | Dose of CD34+ cells per kg of recipient body weight |
| CD3dCD34 | - | Ratio of CD3+ cells to CD34+ cells |
| CD3dkgx10d8 | - | Dose of CD3+ cells per kg of recipient body weight |
| Rbodymass | - | Body mass of the recipient at transplantation time |
| ANCrecovery | - | Time taken for neutrophil recovery, defined as > 0.5 x 10 9 / L |
| PLTrecovery | - | Time taken for platelet recovery, defined as > 50000/mm \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$^3$$\end{document} |
| time_to_aGvHD_III_IV | - | Duration until development of acute graft versus host disease stage III or IV |
| survival_time | - | Observation period (if alive) or time to event (if dead), in days |
| survival_status | - | Status of survival at the end of the observation period |
Pre-processing
Data pre-processing is a crucial step in preparing the “Bone marrow transplant: children” dataset for analysis. This phase involves handling missing values, addressing outliers, and standardizing the data to ensure accurate and reliable results.
Handling Missing Values : In handling missing values within the dataset, a decision has been made to replace them with the value of -1. This choice stems from a deliberate selection of a placeholder that lies outside the expected range of the primary data values, which predominantly consist of 0s and 1s. By substituting missing values with -1, we aim to distinguish them from the actual data points while maintaining a numeric representation that aligns with the binary nature of many features in the dataset [ 32 ].
Feature selection using Particle Swarm Optimization (PSO)
Feature selection is a mandatory step in enhancing the efficiency and performance of ML models, as it aids in identifying the most relevant variables for predictive tasks. In this study, we utilized the PSO for feature selection. PSO is a cutting-edge optimization algorithm that draws inspiration from the natural behavior of bird flocking or fish schooling [ 33 , 34 ]. It aims to investigate a vast feature space and identify the most promising set of features that maximizes a specific criterion (e.g., classification accuracy) [ 35 , 36 ].
The PSO algorithm initializes a population of candidate solutions, referred to as particles, with each particle representing a potential feature subset. The positions and velocities of these particles are randomly initialized within the search space. Each particle in the swarm is considered as a possible feature subset, and the algorithm iteratively refines these subsets to converge towards an optimal solution [ 37 ].
A value between 0 and 1 represents each solution generated. A thresholding mechanism is applied to decide whether a particular feature is included or discarded in the subset. Specifically, a threshold of 0.5 is employed as the decision criterion. Suppose the value associated with a feature in the solution is equal to or greater than 0.5. In that case, the feature is considered for inclusion. Conversely, if the value is below 0.5, the feature is discarded [ 38 ].
During each iteration, particles update the corresponding positions and velocities based on: (1) the personal best-known position ( \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$pBest$$\end{document} ) (cognitive influence) and (2) the global best-known position discovered by any particle in the swarm ( \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$gBest$$\end{document} ) (social influence). The interplay between personal and global influences guides this movement towards superior solutions. The algorithm keeps iterating until a stopping condition is satisfied, such as attaining a maximum number of iterations or finding a sufficiently satisfactory solution. This process ensures that PSO exhaustively investigates the feature space, ultimately identifying the most optimal subset of features [ 39 ].
Regarding our study, let’s denote the position of a particle in the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$D$$\end{document} -dimensional search space as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$X_i = (x_{i1}, x_{i2}, \ldots, x_{iD})$$\end{document} , where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$i$$\end{document} reflects the particle index. The particle velocity is denoted as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$V_i = (v_{i1}, v_{i2}, \ldots, v_{iD})$$\end{document} . The PSO position update equation is presented by Eq. 1 where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$t$$\end{document} is the iteration index, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$j$$\end{document} represents the dimension. The velocity update equation is presented by Eq. 2 , where the inertia weight ( \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$w$$\end{document} ) regulates the balance between the particle’s current velocity and its historical velocity, influencing the exploration and exploitation phases. The acceleration constants ( \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$c_1$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$c_2$$\end{document} ) determine the impact of the particle’s personal best position ( \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$pBest_{ij}$$\end{document} ) and the global best position ( \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$gBest_{ij}$$\end{document} ) on its velocity update. Additionally, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r_1$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r_2$$\end{document} represent random values between 0 and 1, introducing stochasticity to the particles’ movement within the search space. The personal best position ( \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$pBest_{ij}$$\end{document} ) signifies the optimal position a particle has achieved in dimension \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$j$$\end{document} . In contrast, the global best position ( \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$gBest_{ij}$$\end{document} ) is the best-known position across the entire swarm in the same dimension.
1 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$x_{ij}^{t+1} = x_{ij}^t + v_{ij}^{t+1}$$\end{document}
2 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$v_{ij}^{t+1} = \left(\begin{aligned}&w \times v_{ij}^t \\+~& c_1 \times r_1 \times (pBest_{ij} - x_{ij}^t) \\+~& c_2 \times r_2 \times (gBest_{ij} - x_{ij}^t)\end{aligned}\right)$$\end{document}
After the PSO algorithm completes its iterations, the selected features can be extracted from the global best position ( \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$gBest$$\end{document} ), representing the optimal subset identified by the PSO algorithm. These selected features can then be applied in subsequent tasks, such as training ML models or conducting analysis, with the goal of utilizing the subset for improved model performance or interpretability.
Training and optimization
In the “Training and Optimization” phase, a set of ML models is employed, each chosen for its unique strengths and capabilities [ 40 – 42 ]. Among them, SVM optimally separates data points in a high-dimensional space. At the same time, RF utilizes an ensemble of decision trees for robust predictions [ 43 ]. AdaBoost combines weak learners sequentially to create a strong, adaptive model. LightGBM (LGBM) offers efficient training and high performance on large datasets. XGBoost (XGB) is selected for its scalability and accuracy, and Histogram-based Gradient Boosting (HGB) efficiently represents feature value distributions [ 44 ].
Multilayer Perceptron (MLP) learns complex patterns, k-Nearest Neighbors (KNN) relies on proximity for classification, and Decision Tree (DT) provides interpretable decision-making. Gradient Boosting (GB) combines weak learners for powerful predictions, while Extra Trees (ET) introduce additional randomization for diversity. Logistic Regression (LR) is chosen for its simplicity and interpretability, and CatBoost excels in handling categorical features during training [ 45 , 46 ].
Hyperparameters : For the DT classifier, the hyperparameter space encompasses crucial parameters such as maximum depth, the strategy for node splitting (best or random), and the criterion used for the split, either Gini or entropy. SVM utilizes a distinct set of hyperparameters. These include the regularization parameter (C), which is log-normally distributed, the kernel type (linear, radial basis function, RBF, polynomial, or Sigmoid), gamma, and the degree of the polynomial kernel.
The LR hyperparameter space involves the regularization parameter (C) and the choice of solver, either Lib-Linear or LBFGS. RF utilizes the maximum depth, the number of estimators, and the criterion for node splitting (Gini or entropy), which are included.
KNN classifier is characterized by hyperparameters such as the number of neighbors, weighting strategy (uniform or distance-based), algorithm type (Ball Tree, KDTree, or brute force), and the choice of distance metric. LGBM incorporates hyperparameters such as the number of estimators, maximum depth, and the learning rate.
XGV has its hyperparameter space, including the subsample ratio, the estimators number, maximum depth, and learning rate. HGB focuses on maximum depth and learning rate as key hyperparameters in its search space. GB encompasses maximum depth, learning rate, and the estimators number.
AdaBoost’s hyperparameter space is tailored to include parameters such as the estimators number and the learning rate. CatBoost includes the estimators number and early stopping rounds. MLP incorporates hyperparameters like the activation function (logistic, relu, or tanh), learning rate strategy (constant, adaptive, or invscaling), solver choice (LBFGS, SGD, or Adam), and the size of hidden layers. ET includes hyperparameters such as maximum depth, criterion for node splitting (Gini or entropy), and the estimators number.
Bayesian Optimization using Adaptive Tree of Parzen Estimators TPE is a sophisticated approach for hyperparameter optimization, particularly in ML models [ 47 ]. This approach uses the Bayesian optimization technique [ 48 ] and the TPE algorithm to efficiently explore and select the best hyperparameters within a predetermined search space. Bayesian optimization is an emerging technique for sequential model-based optimization [ 49 , 50 ]. It identifies the optimal set of hyperparameters by iteratively constructing a probabilistic model with respect to the objective (or fitness) function. Moreover, it effectively guides the search process. Unlike traditional approaches that exhaustively evaluate all potential hyperparameter configurations, Bayesian optimization strategically explores the most promising regions within the search space.
TPE is a specialized algorithm employed in Bayesian optimization frameworks [ 48 ]. It utilizes two probability distributions to model the objective function: one for favorable configurations (exploitation) and another for unfavorable configurations (exploration). TPE constructs a tree structure to effectively explore the hyperparameter space by progressively dividing it into regions more likely to encompass optimal configurations. Bayesian Optimization with TPE provides various advantages for hyperparameter optimization, including efficiency, robustness, and automatic tuning [ 51 ].
As noted, the core idea behind TPE is to model two probability distributions: one for configurations that lead to an improvement in the objective function ( \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P(x \in A)$$\end{document} ) and another for configurations that do not ( \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P(x \in B)$$\end{document} ). The probability of improvement ( \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P(I)$$\end{document} ) is then used to guide the search towards configurations that are more likely to improve the objective function as shown in Eq. 3 where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$A$$\end{document} is the set of hyperparameter configurations resulting in an improvement, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$B$$\end{document} is the set of all possible hyperparameter configurations [ 52 ].
3 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P(I) = \frac{P(x \in A)}{P(x \in B)}$$\end{document}
To determine \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P(x \in A)$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P(x \in B)$$\end{document} , TPE uses Eqs. 4 and 5 where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\eta$$\end{document} is a constant controlling the trade-off between exploration and exploitation. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p(x \mid I)$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p(x \mid I')$$\end{document} represent the probability density functions of configurations \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$x$$\end{document} given improvement and no improvement, respectively.
4 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P(x \in A) = \eta \times p(x \mid I)$$\end{document}
5 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P(x \in B) = (1 - \eta) \times p(x \mid I')$$\end{document}
The probability density functions are modeled using a kernel density estimation approach. For the given configurations \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$x$$\end{document} , Eq. 6 represents the density estimation where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$k(x, x_i)$$\end{document} is the kernel density function, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\epsilon$$\end{document} is a small constant to avoid division by zero.
6 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p(x \mid I) = \frac{k(x, x_i)}{\sum_{i' \in I} k(x, x_{i'}) + \epsilon}$$\end{document}
TPE adapts and refines the search space dynamically, making it particularly effective in handling complex and high-dimensional optimization problems. The algorithm iteratively proposes new configurations, evaluates their performance, updates the probabilistic models, and refines the search, converging towards the hyperparameter values that optimize the chosen objective function.
Scaling techniques are preprocessing methods employed to standardize or normalize the features of a dataset, guaranteeing their comparability [ 53 , 54 ]. This is of utmost significance in machine learning, as features with varying scales can result in biased or inefficient models, especially for algorithms that heavily rely on distance measures or gradient descent optimization. One key reason for employing scaling techniques is the impact of differing scales on distance-based algorithms [ 55 ]. These algorithms, such as k-nearest neighbors or support vector machines, calculate distances between data points to make predictions or classifications. If the features have varying scales, the distances calculated will be heavily influenced by the features with larger scales, potentially overshadowing the contributions of other features. Scaling mitigates this issue by placing all features equally, ensuring each feature contributes proportionally to the distance calculations.
L1 and L2 normalization are widely-used techniques in scaling or regularizing data, particularly in machine learning, signal processing, and optimization [ 56 , 57 ]. They adjust the values of a vector to meet specific constraints, usually to improve model performance, interpretability, or robustness. L1 normalization (also known as Lasso Regularization) scales a vector so that the sum of its absolute values equals one. L1 regularization employs absolute values, making it less sensitive to outliers and encouraging sparse solutions, which is helpful for feature selection. L2 normalization (Ridge Regularization) scales a vector such that the sum of the squares of its components equals 1. L2 regularization chooses modest, evenly distributed weights, resulting in smoother and more stable models. The squared term in L2 facilitates optimization using gradient-based approaches.
Another critical aspect is the effect of feature scales on gradient descent optimization [ 58 ]. Gradient descent is a widely used optimization algorithm that iteratively adjusts model parameters to minimize a cost function. When features have different scales, the gradients computed for each feature will also differ in magnitude. This can lead to inefficient convergence or slow learning rates, as the algorithm may take longer to find the optimal solution. Scaling addresses this problem by normalizing the features, allowing the algorithm to converge more efficiently and accelerate the learning process.
In this study, the impact of different scaling techniques on model performance is investigated. Various scalers, including L1 normalization, L2 normalization, Maximum absolute scaling, Standard scaling, Min-Max scaling, Max absolute scaling, and Robust scaling, are applied. These techniques are widely used methods in scaling or regularizing data [ 56 , 57 ]. They adjust the values of a vector to meet specific constraints, usually to improve model performance, interpretability, or robustness.
L1 normalization (or Lasso Regularization) scales a vector so that the sum of its absolute values equals one, making it less sensitive to outliers and encouraging sparse solutions, which is helpful for feature selection. L2 normalization (or Ridge Regularization) scales a vector such that the sum of the squares of its components equals 1. It chooses modest, evenly distributed weights, resulting in smoother and more stable models. The squared term in L2 facilitates optimization using gradient-based approaches. Maximum Absolute Scaling, Standard Scaling, Min-Max Scaling, and Max Absolute Scaling offer solutions for normalizing features based on maximum values, Z-score, specific ranges, or global maximum absolute values. Additionally, robust scaling, which utilizes the median and interquartile range, proves effective in handling datasets with outliers.
Performance measurement
Evaluating the ML models’ effectiveness hinges on performance measurement, a critical aspect of the assessment process [ 59 – 61 ]. Many metrics offer insights into different facets of model performance and facilitate a thorough evaluation. Commonly utilized metrics encompass accuracy, precision, recall, specificity, F1 score, and Balanced Accuracy (BAC) [ 62 – 64 ]. They play a pivotal role in quantifying the model’s effectiveness in correctly classifying instances and addressing imbalances within the dataset [ 65 ].
The accuracy, depicted by Eq. 7 , gauges the overall predictions correctness. Precision, articulated in Eq. 8 , concentrates on the accuracy of positive predictions, while recall, as defined in Eq. 9 , underscores the model’s capacity to capture all positive instances. Specificity, outlined in Eq. 10 , evaluates the model’s aptitude for accurately identifying negative instances.
The F1 score, presented in Eq. 11 , strikes a balance between precision and recall, proving valuable in scenarios with uneven class distributions. Balanced Accuracy (BAC), per Eq. 12 , furnishes a well-rounded measure of accuracy, accounting for the performance of both positive and negative classes.
7 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text{Accuracy} = \frac{\text{TN + TP}}{\text{TN + TP + FP + FN}}$$\end{document}
8 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text{Precision} = \frac{\text{TP}}{\text{TP + FP}}$$\end{document}
9 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text{Recall} = \text{Sensitivity} = \frac{\text{TP}}{\text{TP + FN}}$$\end{document}
10 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text{Specificity} = \frac{\text{TN}}{\text{TN + FP}}$$\end{document}
11 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text{F1} = 2 \times \frac{\text{Precision} \times \text{Recall}}{\text{Precision + Recall}}$$\end{document}
12 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text{BAC} = \frac{1}{2} \times \left(\frac{\text{TP}}{\text{TP + FN}} + \frac{\text{TN}}{\text{TN + FP}}\right)$$\end{document}
Local Interpretable Model-Agnostic Explanations (LIME) explanability
The integration of Explainable AI (XAI) techniques is crucial for understanding the predictions made by ML models, particularly in situations where comprehending the reasoning behind these predictions is vital for decision-making, compliance, or trust building [ 66 – 68 ]. LIME is a powerful technique for unraveling ML model predictions that takes on added significance when examining the intricacies of childhood BMT [ 69 , 70 ].
One of the primary benefits of LIME is its model-agnostic nature, which means it can be applied to any black-box ML model without the need for knowledge of its internal behavior. It constructs interpretable surrogate models, such as linear models or decision trees, to simulate the behavior of the black-box model close to the instance that has to be explained. LIME determines which features impact the prediction of the model for a given instance most by looking at the weights that the surrogate model allocated to various features [ 66 – 68 ].
LIME, in its core methodology, involves introducing perturbations around a specific instance, observing the model’s response, and constructing a locally interpretable model to elucidate the prediction rationale. This process is encapsulated by Eq. 13 [ 71 ].
13 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$g = \arg\min_g L(f, g, \pi_{x'}) + \Omega(g)$$\end{document}
The loss function \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$L$$\end{document} is presented in Eq. 14 where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n$$\end{document} represents the number of perturbed instances, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$x_i$$\end{document} denotes the features of the original instance, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$w_i$$\end{document} represents weights associated with each instance, reflecting their importance. This formulation emphasizes the significance of minimizing the absolute differences between the original prediction of the model \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$f(x_i)$$\end{document} and the locally interpretable prediction of the model \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$g(x'_i)$$\end{document} .
14 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$L(f, g, \pi_{x'}) = \sum_{i=1}^{n} w_i \times \left|f(x_i) - g(x'_i)\right|$$\end{document}
That complexity term \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Omega(g)$$\end{document} that penalizes overly complex models is presented in Eq. 15 where is uses the L1 regularization term, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$m$$\end{document} represents the number of features in the interpretable model, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\lambda$$\end{document} is a regularization parameter controlling the strength of the penalty. This term encourages a sparse and simpler locally interpretable model [ 72 ].
15 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Omega(g) = \lambda \times \sum_{j=1}^{m} \left| \frac{\partial g}{\partial x'_j} \right|$$\end{document}
Experiments and discussion
The experiments were conducted using Python on Windows 11, with Anaconda managing packages and environments. They utilized an Intel Core i7 processor, 256 GB of RAM, and an NVIDIA GPU with 8 GB of dedicated memory to ensure sufficient computational power for training and evaluating models. The study employed two cross-validation schemes (5-fold and 10-fold), each repeated over 100 trials to ensure robustness, with performance metrics reported as mean values with a 95% confidence interval, including BAC and recall.
PSO was used for feature selection, while Bayesian Optimization with Adaptive TPE fine-tuned hyperparameters for seven ML algorithms: AdaBoost, CatBoost, ET, LR, MLP, RF, and SVM, all combined using majority voting. Additionally, the research explored the impact of L1 and L2 normalization techniques and utilized LIME for model interpretability. It used the “Bone Marrow Transplant: Children” dataset to evaluate survival prediction in pediatric BMT, aiming to address challenges such as statistical limitations, evolving treatment protocols, and AI model transparency. This is summarized in Table 3 . Table 3 Tabular presentation of the experimental configurations utilized in the current study Parameter Value(s) Software Environment Python on Windows 11, managed by Anaconda Hardware Intel Core i7, 256 GB RAM, NVIDIA GPU (8 GB) Cross-Validation 5-fold and 10-fold Number of Trials 100 trials per setup Dataset “Bone Marrow Transplant: Children” Performance Metrics Mean \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 95% CI (Accuracy, Precision, Recall, etc.) Algorithms AdaBoost, CatBoost, ET, LR, MLP, RF, SVM Hyperparameter Optimization Bayesian Optimization with TPE Feature Selection Particle Swarm Optimization (PSO) Scaling Techniques L1 and L2 normalization
Table 3
Tabular presentation of the experimental configurations utilized in the current study
| Parameter | Value(s) |
| Software Environment | Python on Windows 11, managed by Anaconda |
| Hardware | Intel Core i7, 256 GB RAM, NVIDIA GPU (8 GB) |
| Cross-Validation | 5-fold and 10-fold |
| Number of Trials | 100 trials per setup |
| Dataset | “Bone Marrow Transplant: Children” |
| Performance Metrics | Mean \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 95% CI (Accuracy, Precision, Recall, etc.) |
| Algorithms | AdaBoost, CatBoost, ET, LR, MLP, RF, SVM |
| Hyperparameter Optimization | Bayesian Optimization with TPE |
| Feature Selection | Particle Swarm Optimization (PSO) |
| Scaling Techniques | L1 and L2 normalization |
Preliminary statistical analysis
The analysis conducted (presented in Fig. 8 ) using ANOVA and T-tests reveals significant differences in various factors concerning survival_status. ANOVA results indicate that Donorage, extcGvHD, PLTrecovery, and survival_time exhibit extremely low p-values (p = 0.0000), suggesting a strong association with survival_status. Additionally, Disease, ANCrecovery, and time_to_aGvHD_III_IV also show statistical significance with p-values below 0.05 (p = 0.0175, p = 0.0271, p = 0.0002, respectively). Meanwhile, T-tests further emphasize the impact of variables like RecipientRh, ABOmatch, CMVstatus, Disease, HLAmatch, Antigen, Alel, HLAgrI, Recipientage, Recipientageint, CD34kgx10d6, CD3dCD34, CD3dkgx10d8, Rbodymass, ANCrecovery, PLTrecovery, and time_to_aGvHD_III_IV on survival_status, all displaying very low p-values (p = 0.0000). Fig. 8 Visualization of the ANOVA and T-tests analysis between the features and the target column
The profound insights gained from the statistical analysis highlight the pivotal role of various factors. However, it is crucial to recognize that these findings not only contribute to a deeper understanding of the dataset but also emphasize the importance of incorporating ML models. Utilizing ML techniques allows us to harness the predictive power of these key features, enabling more accurate and informed decision-making. Building a robust ML model on this foundation can enhance our ability to forecast survival_status.
Experimental analysis
Table 4 displays the performance metrics, expressed as mean values with a 95% confidence interval, obtained from applying PSO feature selection within a 5-fold cross-validation framework. The experiments were carried out over 100 trials using the “Bone marrow transplant: children”. The “Best” row shows the performance metrics achieved by the top-performing models, as determined by majority voting. Table 4 Tabular representation of the performance metrics (mean \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 95% CI) showcasing the results of the PSO feature selection conducted within a 5-fold cross-validation framework Approach Accuracy BAC Precision Recall Specificity F1 IoU SVM 95.19% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 95.2% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 96.04% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 95.1% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 95.29% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 95.57% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 91.51% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 RF 95.27% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.001 94.94% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.001 93.15% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0011 98.59% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0014 91.29% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0015 95.79% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0009 91.93% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0017 AdaBoost 95.19% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 95.0% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 94.29% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 97.06% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 92.94% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 95.65% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 91.67% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 LGBM 96.79% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 96.67% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 96.15% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 98.04% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 95.29% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 97.09% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 94.34% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 XGB 94.65% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 94.31% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 92.59% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 98.04% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 90.59% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 95.24% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 90.91% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 HGB 95.19% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 94.9% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 93.46% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 98.04% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 91.76% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 95.69% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 91.74% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 MLP 94.11% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.001 94.17% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0011 95.55% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0015 93.58% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0011 94.75% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0019 94.55% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0009 89.67% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0017 KNN 93.05% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 92.65% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 90.83% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 97.06% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 88.24% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 93.84% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 88.39% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 DT 94.14% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0019 94.02% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.002 93.99% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0029 95.38% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0015 92.66% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0038 94.68% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0017 89.9% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.003 GB 96.26% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 96.18% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 96.12% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 97.06% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 95.29% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 96.59% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 93.4% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 ET 95.23% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.001 95.13% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0011 95.12% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0015 96.21% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.001 94.06% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.002 95.65% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0009 91.68% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0017 LR 97.33% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 97.45% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 98.99% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 96.08% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 98.82% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 97.51% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 95.15% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 CatBoost 95.72% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 95.29% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 92.73% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 100.0% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 90.59% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 96.23% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 92.73% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 Best 98.07% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0008 98.08% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0008 98.45% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0013 98.02% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0003 98.14% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0016 98.23% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0007 96.53% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0013 The experiments were performed over 100 trials using the “Bone marrow transplant: children” dataset
Table 4
Tabular representation of the performance metrics (mean \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 95% CI) showcasing the results of the PSO feature selection conducted within a 5-fold cross-validation framework
The experiments were performed over 100 trials using the “Bone marrow transplant: children” dataset
| Approach | Accuracy | BAC | Precision | Recall | Specificity | F1 | IoU |
| SVM | 95.19% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 95.2% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 96.04% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 95.1% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 95.29% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 95.57% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 91.51% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 |
| RF | 95.27% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.001 | 94.94% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.001 | 93.15% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0011 | 98.59% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0014 | 91.29% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0015 | 95.79% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0009 | 91.93% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0017 |
| AdaBoost | 95.19% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 95.0% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 94.29% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 97.06% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 92.94% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 95.65% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 91.67% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 |
| LGBM | 96.79% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 96.67% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 96.15% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 98.04% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 95.29% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 97.09% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 94.34% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 |
| XGB | 94.65% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 94.31% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 92.59% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 98.04% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 90.59% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 95.24% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 90.91% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 |
| HGB | 95.19% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 94.9% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 93.46% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 98.04% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 91.76% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 95.69% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 91.74% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 |
| MLP | 94.11% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.001 | 94.17% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0011 | 95.55% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0015 | 93.58% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0011 | 94.75% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0019 | 94.55% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0009 | 89.67% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0017 |
| KNN | 93.05% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 92.65% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 90.83% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 97.06% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 88.24% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 93.84% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 88.39% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 |
| DT | 94.14% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0019 | 94.02% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.002 | 93.99% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0029 | 95.38% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0015 | 92.66% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0038 | 94.68% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0017 | 89.9% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.003 |
| GB | 96.26% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 96.18% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 96.12% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 97.06% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 95.29% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 96.59% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 93.4% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 |
| ET | 95.23% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.001 | 95.13% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0011 | 95.12% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0015 | 96.21% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.001 | 94.06% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.002 | 95.65% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0009 | 91.68% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0017 |
| LR | 97.33% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 97.45% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 98.99% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 96.08% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 98.82% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 97.51% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 95.15% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 |
| CatBoost | 95.72% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 95.29% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 92.73% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 100.0% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 90.59% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 96.23% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 92.73% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 |
| Best | 98.07% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0008 | 98.08% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0008 | 98.45% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0013 | 98.02% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0003 | 98.14% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0016 | 98.23% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0007 | 96.53% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0013 |
In the optimal model configuration, as presented in Table 4 , a combination of seven ML algorithms was employed using majority voting: AdaBoost, CatBoost, ET, LR, MLP, RF, and SVM. Each algorithm was fine-tuned with specific hyperparameters to enhance performance. For instance, the MLP model utilized an activation function of ReLU, a hidden layer size of 128, a constant learning rate, and the Adam solver. CatBoost incorporated early stopping with 50 rounds and 76 estimators. The LR model had a regularization parameter (C) of 2.84768472512 and used the LibLinear solver. RF adopted entropy as the criterion, had a maximum depth of 17, and used 47 estimators. SVM’s hyperparameters included a regularization parameter (C) of 0.9097850578, a degree of 4, Auto for gamma, and an RBF kernel. ET employed Gini as the criterion, a max depth of 12, and 94 estimators. AdaBoost tuned its learning rate to 0.80355585192 and used 68 estimators. The number of features considered for each model varied, with MLP using 12, CatBoost with 23, LR with 21, RF with 19, SVM with 19, ET with 16, and AdaBoost with 18 features. Additionally, different scalers were applied, such as Robust for MLP and CatBoost, STD for LR and SVM, MaxAbs for RF, and L2 for ET and AdaBoost.
The “Best” approach in Table 4 summarizes the superior performance metrics achieved by the most effective ML algorithm in the context of PSO feature selection applied to the “Bone marrow transplant: children” dataset. It outperforms others in terms of accuracy (98.07% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0008), BAC (98.08% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0008), precision (98.45% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0013), recall (98.02% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0003), specificity (98.14% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0016), F1 score (98.23% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0007), and IoU (96.53% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0013).
Table 5 displays the performance metrics, expressed as mean values with a 95% confidence interval, obtained from applying PSO feature selection within a 10-fold cross-validation framework. The experiments were carried out over 100 trials using the “Bone marrow transplant: children”. The “Best” row shows the performance metrics achieved by the top-performing models, as determined by majority voting. Table 5 Tabular representation of the performance metrics (mean \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 95% CI) showcasing the results of the PSO feature selection conducted within a 10-fold cross-validation framework Approach Accuracy BAC Precision Recall Specificity F1 IoU SVM 94.12% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 93.92% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 93.33% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 96.08% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 91.76% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 94.69% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 89.91% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 RF 94.59% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0015 94.39% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0015 93.64% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0013 96.66% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0023 92.12% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0017 95.12% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0014 90.7% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0025 AdaBoost 94.65% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 94.41% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 93.4% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 97.06% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 91.76% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 95.19% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 90.83% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 LGBM 97.33% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 97.16% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 96.19% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 99.02% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 95.29% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 97.58% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 95.28% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 XGB 96.79% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 96.57% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 95.28% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 99.02% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 94.12% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 97.12% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 94.39% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 HGB 94.12% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 93.73% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 91.74% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 98.04% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 89.41% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 94.79% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 90.09% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 MLP 95.69% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0002 95.66% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0002 96.08% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 96.02% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0003 95.29% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 96.05% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0002 92.4% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0003 KNN 92.51% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 92.06% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 90.0% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 97.06% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 87.06% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 93.4% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 87.61% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 DT 94.95% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0007 94.84% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0008 94.75% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0012 96.08% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 93.6% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0015 95.41% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0006 91.22% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0011 GB 93.8% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0009 93.62% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.001 93.22% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0012 95.6% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0011 91.65% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0016 94.39% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0008 89.38% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0015 ET 94.36% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0012 93.84% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0012 90.95% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0015 99.57% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0011 88.11% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0021 95.06% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.001 90.6% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0019 LR 97.33% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 97.35% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 98.02% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 97.06% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 97.65% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 97.54% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 95.19% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 CatBoost 94.12% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 93.63% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 90.99% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 99.02% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 88.24% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 94.84% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 90.18% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 Best 97.86% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 97.75% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 97.12% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 99.02% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 96.47% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 98.06% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 96.19% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 The experiments were performed over 100 trials using the “Bone marrow transplant: children” dataset
Table 5
Tabular representation of the performance metrics (mean \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 95% CI) showcasing the results of the PSO feature selection conducted within a 10-fold cross-validation framework
The experiments were performed over 100 trials using the “Bone marrow transplant: children” dataset
| Approach | Accuracy | BAC | Precision | Recall | Specificity | F1 | IoU |
| SVM | 94.12% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 93.92% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 93.33% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 96.08% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 91.76% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 94.69% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 89.91% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 |
| RF | 94.59% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0015 | 94.39% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0015 | 93.64% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0013 | 96.66% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0023 | 92.12% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0017 | 95.12% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0014 | 90.7% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0025 |
| AdaBoost | 94.65% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 94.41% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 93.4% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 97.06% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 91.76% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 95.19% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 90.83% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 |
| LGBM | 97.33% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 97.16% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 96.19% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 99.02% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 95.29% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 97.58% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 95.28% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 |
| XGB | 96.79% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 96.57% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 95.28% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 99.02% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 94.12% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 97.12% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 94.39% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 |
| HGB | 94.12% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 93.73% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 91.74% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 98.04% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 89.41% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 94.79% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 90.09% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 |
| MLP | 95.69% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0002 | 95.66% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0002 | 96.08% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 96.02% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0003 | 95.29% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 96.05% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0002 | 92.4% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0003 |
| KNN | 92.51% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 92.06% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 90.0% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 97.06% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 87.06% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 93.4% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 87.61% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 |
| DT | 94.95% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0007 | 94.84% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0008 | 94.75% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0012 | 96.08% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 93.6% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0015 | 95.41% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0006 | 91.22% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0011 |
| GB | 93.8% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0009 | 93.62% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.001 | 93.22% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0012 | 95.6% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0011 | 91.65% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0016 | 94.39% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0008 | 89.38% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0015 |
| ET | 94.36% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0012 | 93.84% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0012 | 90.95% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0015 | 99.57% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0011 | 88.11% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0021 | 95.06% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.001 | 90.6% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0019 |
| LR | 97.33% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 97.35% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 98.02% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 97.06% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 97.65% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 97.54% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 95.19% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 |
| CatBoost | 94.12% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 93.63% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 90.99% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 99.02% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 88.24% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 94.84% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 90.18% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 |
| Best | 97.86% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 97.75% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 97.12% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 99.02% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 96.47% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 98.06% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 | 96.19% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0 |
In crafting the optimal model configuration, as detailed in Table 5 , a synergistic approach involving seven ML algorithms, namely AdaBoost, CatBoost, ET, LR, MLP, RF, and SVM, was employed. Each algorithm underwent meticulous fine-tuning with specific hyperparameters tailored to enhance its performance. Noteworthy specifics include AdaBoost’s learning rate of 0.80355585192 and 68 estimators, CatBoost’s early stopping with 50 rounds and 76 estimators, ET’s use of Gini as the criterion, LR’s regularization parameter (C) of 2.84768472512, MLP’s ReLU activation function and 128 hidden layer size, RF’s adoption of entropy as the criterion with a maximum depth of 17 and 47 estimators, and SVM’s hyperparameters, including a regularization parameter (C) of 0.9097850578 and a degree of 4. The number of features considered varied across models, with MLP utilizing 20, CatBoost with 23, LR with 21, RF with 19, SVM and ET with 19, and AdaBoost with 18 features. Additionally, distinct scalers, such as Robust for MLP and CatBoost, Standard (STD) for LR and SVM, MaxAbs for RF, and Min-Max for MLP, were judiciously applied.
The “Best” approach in Table 5 summarizes the superior performance metrics achieved by the most effective ML algorithm in the context of PSO feature selection applied to the utilized dataset. This top-performing approach demonstrates superior outcomes in terms of accuracy (97.86% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0), BAC (97.75% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0), precision (97.12% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0), recall (99.02% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0), specificity (96.47% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0), F1 score (98.06% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0), and IoU (96.19% \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pm$$\end{document} 0.0).
Figure 9 provides a visual depiction of feature importance across various ML algorithms applied to the utilized dataset. The visualization is divided into two panels: the left panel shows feature importance for 5-fold cross-validation, while the right panel presents the results for 10-fold cross-validation. This comprehensive display enables a comparative analysis of how different algorithms rank features in the context of both cross-validation scenarios. Insights derived from the feature importance analysis highlight the relative significance of each feature in influencing the predictive performance of the algorithms on the dataset. Fig. 9 Visualization of the feature importances for all features across the applied ML algorithms using the “Bone marrow transplant: children”. ( a ) illustrates feature importances for five-fold, while ( b ) depicts them for ten-fold
The detailed analysis of feature importance reveals the consistent significance of key measures such as survival_time, Relapse, CD34kgx10d6, and extcGvHD across both 5-fold and 10-fold cross-validation scenarios. This highlights the critical role these factors play in determining the outcomes of childhood BMT. Such insights are highly valuable for medical professionals, directing their attention to the most impactful aspects when evaluating and monitoring the progression of hematological disorders in pediatric BMT patients. By focusing on these prioritized features, physicians can enhance the precision and accuracy of their assessments, paving the way for more effective and targeted interventions for children facing these complex medical challenges.
Figure 10 presents a visual comparison of ROC curves, highlighting the performance of leading ML algorithms on the utilized dataset. The left panel of the plot displays the ROC curves for 5-fold cross-validation. In contrast, the right panel illustrates the corresponding curves for 10-fold cross-validation. This graphical representation enables a comprehensive assessment of the algorithms’ performance, emphasizing their sensitivity and specificity across different cross-validation settings. Fig. 10 Visualization of the ROC curves comparing the top ML algorithms with the “Bone marrow transplant: children”. ( a ) shows the ROC curve for five-folds, while ( b ) shows it for ten-folds
Clinical integration and deployment feasibility
To ensure the practical significance and translational potential of the proposed CAD framework for pediatric BMT, a comprehensive strategy is presented for integration into real-world clinical workflows. This includes a detailed technical deployment architecture, specifies the necessary data input requirements, and outlines interpretability mechanisms aimed at enhancing informed clinical decision-making.
System architecture and deployment strategy
The framework is designed using a modular architecture to support flexible deployment in diverse clinical environments. It consists of the following components: Backend AI Engine: Encapsulates the feature selection module (PSO-based), classifier ensemble, and explainability tools. Data Integration Layer: Interfaces with Electronic Health Record (EHR) systems via standard protocols (e.g., HL7, FHIR) to fetch patient data in real-time. Data Integration Layer: Interfaces with Electronic Health Record (EHR) systems via standard protocols (e.g., HL7, FHIR) to fetch patient data in real-time.
Deployment options include standalone on-premises servers or integration into cloud-based hospital IT infrastructure, with compliance to healthcare regulations such as HIPAA and GDPR.
Clinical data requirements
The framework requires routinely collected data that is generally available in pediatric BMT centers: Demographic information (e.g., age, gender). Clinical history (e.g., disease type, prior treatments). Laboratory results (e.g., WBC count, platelet count). Transplant-specific indicators (e.g., engraftment days, GVHD occurrence).
Robust pre-processing and missing data imputation strategies ensure the framework remains functional even when some data fields are incomplete.
Decision thresholds and clinical utility
The system supports customizable classification thresholds to adapt to the clinical context. For instance, clinicians may prioritize high sensitivity to avoid missing high-risk patients or opt for high specificity in resource-limited settings. The risk output is categorized into three tiers (Low Risk, Moderate Risk, and High Risk) with associated confidence scores and explainable feature attributions. Furthermore, the integration of XAI techniques allows clinicians to: Understand the rationale behind predictions. Explore patient-specific risk drivers. Build trust in model recommendations.
This interpretability is critical for clinician acceptance and for supporting shared decision-making with patients and families.
Conclusions and future work
Guided by the insights from our bibliometric study, which traced the evolution of AI and ML applications in pediatric BMT from 2015 to 2023, this research introduces a state-of-the-art CAD framework. The bibliometric analysis highlighted the growing integration of AI and ML in this field and identified key areas for innovation, laying the groundwork for our proposal. This research introduces a groundbreaking CAD framework for pediatric BMT, marking a significant leap in the application of AI and ML within this critical medical domain. Key contributions include the development of the CAD framework, using PSO for precise feature selection, and implementing Bayesian Optimization with Adaptive TPE for optimal model performance tuning. Our comprehensive examination of data scaling techniques has yielded valuable insights for medical data pre-processing, enhancing model accuracy. Besides, integrating LIME has bridged the gap between AI complexity and clinical application, ensuring model transparency and fostering clinician trust. The ensemble of finely-tuned ML algorithms within our CAD framework not only achieves outstanding performance metrics but also provides interpretable insights into the classification process, marking a significant contribution to medical AI and a robust tool for enhancing decision-making in treating childhood BMT. Future research directions include integrating real-time data analysis for dynamic updates during BMT, exploring federated learning to utilize decentralized data while ensuring privacy, and enhancing the interpretability of AI models to improve clinician trust and understanding in AI-driven diagnostics. These advancements aim to refine the CAD framework further and expand the potential of AI applications in pediatric BMT.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Yousry AbdulAzeem, Hanaa ZainEldin, Hossam Magdy Balaha, Amna Bamaqa, Rasha F. El-Agamy, Hanaa A. Sayed and Mostafa A. Elhosseini contributed equally to this work.
Acknowledgements
Not applicable.
Author contributions
Conceptualization, A.B, H.M.B and M.B; Methodology, H.M.B, Y.A, M.A.E. and M.B; Software, H.M.B, H.Z, and Y.A; Validation, A.B, R.F and H.A; Formal analysis, A.B, H.A.,R.F and H.Z; Investigation, H.M.B, A.B, R.A and H.F; Resources, M.A.E, H.M.B, Y.A, H.M.B and M.A.E; Data curation, H.M.B, Y.A, H.Z,R.F and H.A; Writing (original draft), H.M.B, A.B, H.Z R.F and H.A; Writing (review & editing), Y.A, M.B, M.A.E; Visualization, Y.A, M.B and M.A.E; Supervision, M.B; Project administration, M.A.E; Funding acquisition, M.A.E.
Funding
Not applicable.
Data availability
The data that support the findings of this study are available at: 10.24432/C5NP6Z.
Declarations
Ethics approval and consent to participate
Not applicable.
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Abbreviations
Aplastic anemia
Artificial intelligence
Balanced Accuracy
British Bone Marrow Registry
Boosted decision trees
Bone Marrow
Bone marrow transplantation
Computer-Aided Diagnosis
Convolutional Neural Networks
Deep learning
Decision Tree
Feedforward Neural Networks
Gradient Boosting
Grid search cross-validation
Human leukocyte antigen
Hematopoietic Stem Cell Transplantation
Local Interpretable Model-Agnostic Explanations
Logistic Regression
Myelodysplastic syndromes
Machine learning
Mud Ring Algorithm
Magnetic resonance imaging
Naive Bayes
Particle Swarm Optimization
Random Forest
Root-mean-square deviation
Recurrent Neural Networks
Receiver Operating Characteristic
Regions of interest
Severe combined immunodeficiency
Shapley additive values
Support vector machine
Adaptive Tree of Parzen Estimators
Explainable artificial intelligence
References
1. Foucar K Reichard K Czuchlewski D Bone Marrow Pathology. 2019 American Society for Clinical Pathology Press Foucar K, Reichard K, Czuchlewski D. Bone Marrow Pathology. American Society for Clinical Pathology Press; 2019.
2. Meyerson H, Sundaram S, Lazarus HM. Bone marrow structure and marrow aspiration, biopsy, and collection for therapeutic intent procedures. Concise guide to Hematol. 2019;233–52.
3. Park B Yoo KH Kim C Hematopoietic stem cell expansion and generation: the ways to make a breakthrough Blood Res 2015 50 4 194 10.5045/br.2015.50.4.194 26770947 PMC4705045 Park B, Yoo KH, Kim C. Hematopoietic stem cell expansion and generation: the ways to make a breakthrough. Blood Res. 2015;50(4):194. 26770947 10.5045/br.2015.50.4.194 PMC4705045
4. Nabarrete JM, Pereira AZ, Garófolo A, Seber A, Venancio AM, Grecco CES, Bonfim CMS, Nakamura CH, Fernandes D, Campos DJ, et al. Brazilian nutritional consensus in hematopoietic stem cell transplantation: children and adolescents. Einstein (São Paulo). 2021;19:5254. 10.31744/einstein_journal/2021AE5254 PMC8664291 34909973
5. Di Giuseppe G Thacker N Schechter T Pole JD Anxiety, depression, and mental health-related quality of life in survivors of pediatric allogeneic hematopoietic stem cell transplantation: a systematic review Bone Marrow Transplant 2020 55 7 1240 54 10.1038/s41409-020-0782-z 31980743 Di Giuseppe G, Thacker N, Schechter T, Pole JD. Anxiety, depression, and mental health-related quality of life in survivors of pediatric allogeneic hematopoietic stem cell transplantation: a systematic review. Bone Marrow Transplant. 2020;55(7):1240–54. 31980743 10.1038/s41409-020-0782-z
6. Li Y Zhou F Efficacy of bone marrow transplantation in treating acute myeloid leukemia: a systematic review and meta-analysis Am J Transl Res 2023 15 1 1 36777836 PMC9908463 Li Y, Zhou F. Efficacy of bone marrow transplantation in treating acute myeloid leukemia: a systematic review and meta-analysis. Am J Transl Res. 2023;15(1):1. 36777836 PMC9908463
7. Ryou H Lomas O Theissen H Thomas E Rittscher J Royston D Quantitative interpretation of bone marrow biopsies in mpn—what’s the point in a molecular age? Br J Haematol 2023 203 4 523 35 10.1111/bjh.19154 37858962 PMC10952168 Ryou H, Lomas O, Theissen H, Thomas E, Rittscher J, Royston D. Quantitative interpretation of bone marrow biopsies in mpn—what’s the point in a molecular age?. Br J Haematol. 2023;203(4):523–35. 37858962 10.1111/bjh.19154 PMC10952168
8. Kuchakzadeh F Ai J Ebrahimi-Barough S Tissue engineering and stem cell-based therapeutic strategies for premature ovarian insufficiency Regenerative Ther 2024 25 10 23 10.1016/j.reth.2023.11.007 PMC10724490 38108045 Kuchakzadeh F, Ai J, Ebrahimi-Barough S. Tissue engineering and stem cell-based therapeutic strategies for premature ovarian insufficiency. Regenerative Ther. 2024;25:10–23. 10.1016/j.reth.2023.11.007 PMC10724490 38108045
9. Bogucki J Tuszyńska-Bogucka W “be the match”’. predictors of decisions concerning registration as a potential bone marrow donor—a psycho-socio-demographic study Int J Environ Res Pub Health and Public Health 2023 20 11 5993 10.3390/ijerph20115993 PMC10252417 37297597 Bogucki J, Tuszyńska-Bogucka W. “be the match”. predictors of decisions concerning registration as a potential bone marrow donor—a psycho-socio-demographic study. Int J Environ Res Pub Health and Public Health. 2023;20(11):5993. 10.3390/ijerph20115993 PMC10252417 37297597
10. Zhou Y Smith J Keerthi D Li C Sun Y Mothi SS Shyr DC Spitzer B Harris A Chatterjee A Longitudinal clinical data improve survival prediction after hematopoietic cell transplantation using machine learning Blood Adv 2024 8 3 686 98 10.1182/bloodadvances.2023011752 37991991 PMC10844815 Zhou Y, Smith J, Keerthi D, Li C, Sun Y, Mothi SS, Shyr DC, Spitzer B, Harris A, Chatterjee A, et al. Longitudinal clinical data improve survival prediction after hematopoietic cell transplantation using machine learning. Blood Adv. 2024;8(3):686–98. 37991991 10.1182/bloodadvances.2023011752 PMC10844815
11. Murphy JD Cooke KR Symons HJ VanGraafeiland B Enteral nutrition optimization program for children undergoing blood & marrow transplantation: A quality improvement project J Pediatr Nurs 2024 74 61 68 10.1016/j.pedn.2023.11.015 38000117 Murphy JD, Cooke KR, Symons HJ, VanGraafeiland B. Enteral nutrition optimization program for children undergoing blood & marrow transplantation: A quality improvement project. J Pediatr Nurs. 2024;74:61–68. 38000117 10.1016/j.pedn.2023.11.015
12. Dei D, Lambri N, Crespi L, Brioso RC, Loiacono D, Clerici E, Bellu L, De Philippis C, Navarria P, Bramanti S, et al. Deep learning and atlas-based models to streamline the segmentation workflow of total marrow and lymphoid irradiation. La radiologia medica. 2024;1–9. 10.1007/s11547-024-01760-8 38308062
13. Booker JK Diagnostic Molecular Pathology 2024 Elsevier 617 22 Booker JK: Molecular testing in the assessment of bone marrow transplant engraftment. In: Diagnostic Molecular Pathology. Elsevier, 2024. p. 617–22.
14. Chadaga K Prabhu S Sampathila N Chadaga R A machine learning and explainable artificial intelligence approach for predicting the efficacy of hematopoietic stem cell transplant in pediatric patients Healthcare Analytics 2023 3 100170 10.1016/j.health.2023.100170 Chadaga K, Prabhu S, Sampathila N, Chadaga R. A machine learning and explainable artificial intelligence approach for predicting the efficacy of hematopoietic stem cell transplant in pediatric patients. Healthcare Analytics. 2023;3:100170.
15. Li Y Masiliune A Winstone D Gasieniec L Wong P Lin H Pawson R Parkes G Hadley A Predicting the availability of hematopoietic stem cell donors using machine learning Biol of Blood and Marrow Transplant 2020 26 8 1406 13 10.1016/j.bbmt.2020.03.026 32413415 Li Y, Masiliune A, Winstone D, Gasieniec L, Wong P, Lin H, Pawson R, Parkes G, Hadley A. Predicting the availability of hematopoietic stem cell donors using machine learning. Biol of Blood and Marrow Transplant. 2020;26(8):1406–13. 32413415 10.1016/j.bbmt.2020.03.026
16. Marvin G, Alam MGR. Explainable computational pathology for survival prediction in hematologic pediatric patients. 2022 IEEE Asia-Pacific Conference on Computer Science and Data Engineering (CSDE). 2022. 1–6. IEEE.
17. El Bakrawy LM Bailek N Abualigah L Urooj S Desuky AS Feature selection based on mud ring algorithm for improving survival prediction of children undergoing hematopoietic stem-cell transplantation Mathematics 2022 10 22 4197 10.3390/math10224197 El Bakrawy LM, Bailek N, Abualigah L, Urooj S, Desuky AS. Feature selection based on mud ring algorithm for improving survival prediction of children undergoing hematopoietic stem-cell transplantation. Mathematics. 2022;10(22):4197.
18. Zubedi E, Deng Z, Ali Q, Khan A, Shaikh A, Reshan MS, Sulaiman A, Alshahrani H. A machine learning and deep learning-based integrated multi-omics technique for leukemia prediction. Heliyon. 2024. 10.1016/j.heliyon.2024.e25369 PMC10862685 38352790
19. Gupta V Braun TM Chowdhury M Tewari M Choi SW A systematic review of machine learning techniques in hematopoietic stem cell transplantation (hsct) Sensors 2020 20 21 6100 10.3390/s20216100 33120974 PMC7663237 Gupta V, Braun TM, Chowdhury M, Tewari M, Choi SW. A systematic review of machine learning techniques in hematopoietic stem cell transplantation (hsct). Sensors. 2020;20(21):6100. 33120974 10.3390/s20216100 PMC7663237
20. Ratul IJ, Wani UH, Nishat MM, Al-Monsur A, Ar-Rafi AM, Faisal F, Kabir MR, et al. Survival prediction of children undergoing hematopoietic stem cell transplantation using different machine learning classifiers by performing chi-square test and hyperparameter optimization: a retrospective analysis. Comput Math Method M. 2022. 10.1155/2022/9391136 PMC9527434 36199778
21. Javanmehr N, Moayedi H, Farokhnia F. Pediatric hsct survival rates detection through the application of different ann model optimizers, bsa, vsa, and goa. 2023.
22. Sharen H, Narayanan S, Jawahar M, Anbarasi LJ, Raj BE. Uses and trends in hematopoietic cell transplantation in the united states of america: A machine learning review. 2021 5th International Conference on Electrical, Electronics, Communication, Computer Technologies and Optimization Techniques (ICEECCOT). IEEE; 2021, pp. 523–28.
23. Su J Wang Y Zhang J Niu S Han J Xing L Song J Roi-bmc-dnnet: An efficient automatic analysis model of whole-slide scanned bone marrow aspirate images for the diagnosis of hematological disorders Biomed Signal Process Control 2023 86 105243 10.1016/j.bspc.2023.105243 Su J, Wang Y, Zhang J, Niu S, Han J, Xing L, Song J. Roi-bmc-dnnet: An efficient automatic analysis model of whole-slide scanned bone marrow aspirate images for the diagnosis of hematological disorders. Biomed Signal Process Control. 2023;86:105243.
24. Xiang P Wu X Zeng Z Lin Z Guo Y Ma X Lin J Wang W Quantitative analysis of pelvic bone marrow fat using an mri-based machine learning method for distinguishing aplastic anaemia from myelodysplastic syndromes Clin Radiol 2023 78 6 463 68 10.1016/j.crad.2023.02.012 36977621 Xiang P, Wu X, Zeng Z, Lin Z, Guo Y, Ma X, Lin J, Wang W. Quantitative analysis of pelvic bone marrow fat using an mri-based machine learning method for distinguishing aplastic anaemia from myelodysplastic syndromes. Clin Radiol. 2023;78(6):463–68. 10.1016/j.crad.2023.02.012 36977621
25. Gourisaria MK, Patel AV, Chatterjee R, Sahoo B. Predicting the survival status of patient after bone marrow transplant using fuzzy discernibility matrix. 2022 OPJU International Technology Conference on Emerging Technologies for Sustainable Development (OTCON). IEEE; 2023, pp. 1–6.
26. Zhou Y Smith J Keerthi D Sarvode Mothi S Shyr DC Spitzer B Chatterjee A Chatterjee S Naik S Tang L Machine learning approaches incorporating high-dimensional longitudinal data improve the prediction of survival after allogeneic hematopoietic cell transplantation Blood 2022 140 Supplement 1 322 23 10.1182/blood-2022-162890 Zhou Y, Smith J, Keerthi D, Sarvode Mothi S, Shyr DC, Spitzer B, Chatterjee A, Chatterjee S, Naik S, Tang L, et al. Machine learning approaches incorporating high-dimensional longitudinal data improve the prediction of survival after allogeneic hematopoietic cell transplantation. Blood. 2022;140(Supplement 1):322–23.
27. Kockwelp J Thiele S Bartsch J Haalck L Gromoll J Schlatt S Exeler R Bleckmann A Lenz G Wolf S Deep learning predicts therapy-relevant genetics in acute myeloid leukemia from pappenheim-stained bone marrow smears Blood Adv 2024 8 1 70 79 10.1182/bloodadvances.2023011076 37967385 PMC10787267 Kockwelp J, Thiele S, Bartsch J, Haalck L, Gromoll J, Schlatt S, Exeler R, Bleckmann A, Lenz G, Wolf S, et al. Deep learning predicts therapy-relevant genetics in acute myeloid leukemia from pappenheim-stained bone marrow smears. Blood Adv. 2024;8(1):70–79. 37967385 10.1182/bloodadvances.2023011076 PMC10787267
28. Taati B Snoek J Aleman D Ghavamzadeh A Data mining in bone marrow transplant records to identify patients with high odds of survival IEEE J Biomed and Health Inf 2013 18 1 21 27 10.1109/JBHI.2013.2274733 24403400 Taati B, Snoek J, Aleman D, Ghavamzadeh A. Data mining in bone marrow transplant records to identify patients with high odds of survival. IEEE J Biomed and Health Inf. 2013;18(1):21–27. 10.1109/JBHI.2013.2274733 24403400
29. Wang W Luo M Guo P Wei Y Tan Y Shi H Artificial intelligence-assisted diagnosis of hematologic diseases based on bone marrow smears using deep neural networks Comput Met Programs Biomed 2023 231 107343 10.1016/j.cmpb.2023.107343 36821974 Wang W, Luo M, Guo P, Wei Y, Tan Y, Shi H. Artificial intelligence-assisted diagnosis of hematologic diseases based on bone marrow smears using deep neural networks. Comput Met Programs Biomed. 2023;231:107343. 10.1016/j.cmpb.2023.107343 36821974
30. Pinkel D Bone marrow transplantation in children J Sport Hist of pediatrics 1993 122 3 331 41 10.1016/s0022-3476(05)83413-4 8441084 Pinkel D. Bone marrow transplantation in children. J Sport Hist of pediatrics. 1993;122(3):331–41. 10.1016/s0022-3476(05)83413-4 8441084
31. Sikora M Wróbel Ł Gudyś A Bone marrow transplant: children UCI Mach Learn Repository. 2020 10.24432/C5NP6Z Sikora M, Wróbel Ł, Gudyś A. Bone marrow transplant: children. UCI Mach Learn Repository. 2020. 10.24432/C5NP6Z.
32. Saar-Tsechansky M, Provost F. Handling missing values when applying classification models. 2007.
33. Parouha RP Verma P A systematic overview of developments in differential evolution and particle swarm optimization with their advanced suggestion Appl Intell 2022 52 9 10448 92 10.1007/s10489-021-02803-7 Parouha RP, Verma P. A systematic overview of developments in differential evolution and particle swarm optimization with their advanced suggestion. Appl Intell. 2022;52(9):10448–92.
34. Gad AG. Particle swarm optimization algorithm and its applications: a systematic review. Archives of computational methods in engineering. 2022;29(5):2531–61.
35. Liu Y Wang G Chen H Dong H Zhu X Wang S An improved particle swarm optimization for feature selection J Bionic Eng 2011 8 2 191 200 10.1016/S1672-6529(11)60020-6 Liu Y, Wang G, Chen H, Dong H, Zhu X, Wang S. An improved particle swarm optimization for feature selection. J Bionic Eng. 2011;8(2):191–200.
36. Wang D Tan D Liu L Particle swarm optimization algorithm: an overview Soft Comput 2018 22 387 408 10.1007/s00500-016-2474-6 Wang D, Tan D, Liu L. Particle swarm optimization algorithm: an overview. Soft Comput. 2018;22:387–408.
37. Lin S-W Ying K-C Chen S-C Lee Z-J Particle swarm optimization for parameter determination and feature selection of support vector machines Expert Syst with applications 2008 35 4 1817 24 10.1016/j.eswa.2007.08.088 Lin S-W, Ying K-C, Chen S-C, Lee Z-J. Particle swarm optimization for parameter determination and feature selection of support vector machines. Expert Syst with applications. 2008;35(4):1817–24.
38. Marini F Walczak B Particle swarm optimization (pso). a tutorial Chemom and Intell Lab Syst 2015 149 153 65 10.1016/j.chemolab.2015.08.020 Marini F, Walczak B. Particle swarm optimization (pso). a tutorial. Chemom and Intell Lab Syst. 2015;149:153–65.
39. Wang X Yang J Teng X Xia W Jensen R Feature selection based on rough sets and particle swarm optimization Pattern Recognit Lett 2007 28 4 459 71 10.1016/j.patrec.2006.09.003 Wang X, Yang J, Teng X, Xia W, Jensen R. Feature selection based on rough sets and particle swarm optimization. Pattern Recognit Lett. 2007;28(4):459–71.
40. Song C, Ristenpart T, Shmatikov V. Machine learning models that remember too much. Proceedings of the 2017 ACM SIGSAC Conference on Computer and Communications Security. 2017, pp. 587–601.
41. Ataş PK A novel harris hawks optimization-based clustering method for elucidating genetic associations in osteoarthritis and diverse cancer types Comput Biol Med 2025 193 110343 10.1016/j.compbiomed.2025.110343 40412087 Ataş PK. A novel harris hawks optimization-based clustering method for elucidating genetic associations in osteoarthritis and diverse cancer types. Comput Biol Med. 2025;193:110343. 10.1016/j.compbiomed.2025.110343. 40412087 10.1016/j.compbiomed.2025.110343
42. Karadayı Ataş P Exploring the molecular interaction of pcos and endometrial carcinoma through novel hyperparameter-optimized ensemble clustering approaches Mathematics 2024 12 2 295 10.3390/math12020295 Karadayı Ataş P. Exploring the molecular interaction of pcos and endometrial carcinoma through novel hyperparameter-optimized ensemble clustering approaches. Mathematics. 2024;12(2):295. 10.3390/math12020295.
43. Chen P-HC Liu Y Peng L How to develop machine learning models for healthcare Nat Mater 2019 18 5 410 14 10.1038/s41563-019-0345-0 31000806 Chen P-HC, Liu Y, Peng L. How to develop machine learning models for healthcare. Nat Mater. 2019;18(5):410–14. 31000806 10.1038/s41563-019-0345-0
44. Sullivan E. Understanding from machine learning models. Br J Phil Sci. 2022.
45. Carbonell JG, Michalski RS, Mitchell. T.M.: An overview of machine learning. Mach Learn. 1983;3–23.
46. Abd El-Khalek AA Balaha HM Alghamdi NS Ghazal M Khalil AT Abo- Elsoud MEA El-Baz A A concentrated machine learning-based classification system for age-related macular degeneration (amd) diagnosis using fundus images Sci Rep 2024 14 1 2434 10.1038/s41598-024-52131-2 38287062 PMC10825213 Abd El-Khalek AA, Balaha HM, Alghamdi NS, Ghazal M, Khalil AT, Abo- Elsoud MEA, El-Baz A. A concentrated machine learning-based classification system for age-related macular degeneration (amd) diagnosis using fundus images. Sci Rep. 2024;14(1):2434. 38287062 10.1038/s41598-024-52131-2 PMC10825213
47. Tao S Peng P Li Y Sun H Li Q Wang H Supervised contrastive representation learning with tree-structured parzen estimator bayesian optimization for imbalanced tabular data Expert Syst with applications 2024 237 121294 10.1016/j.eswa.2023.121294 Tao S, Peng P, Li Y, Sun H, Li Q, Wang H. Supervised contrastive representation learning with tree-structured parzen estimator bayesian optimization for imbalanced tabular data. Expert Syst with applications. 2024;237:121294.
48. Omotehinwa TO, Oyewola DO, Dada. E.G.: A light gradient-boosting machine algorithm with tree-structured parzen estimator for breast cancer diagnosis. Healthcare Analytics. 2023;100218.
49. Garnett R Bayesian Optimization. 2023 Cambridge University Press Garnett R. Bayesian Optimization. Cambridge University Press; 2023.
50. Wang X Jin Y Schmitt S Olhofer M Recent advances in bayesian optimization ACM Comput Surv 2023 55 13s 1 36 10.1145/3582270 Wang X, Jin Y, Schmitt S, Olhofer M. Recent advances in bayesian optimization. ACM Comput Surv. 2023;55(13s):1–36.
51. Ozaki Y, Tanigaki Y, Watanabe S, Onishi M. Multiobjective tree-structured parzen estimator for computationally expensive optimization problems. Proceedings of the 2020 Genetic and Evolutionary Computation Conference. 2020, pp. 533–41.
52. Rong G Li K Su Y Tong Z Liu X Zhang J Zhang Y Li T Comparison of tree-structured parzen estimator optimization in three typical neural network models for landslide susceptibility assessment Remote Sens 2021 13 22 4694 10.3390/rs13224694 Rong G, Li K, Su Y, Tong Z, Liu X, Zhang J, Zhang Y, Li T. Comparison of tree-structured parzen estimator optimization in three typical neural network models for landslide susceptibility assessment. Remote Sens. 2021;13(22):4694.
53. Ghosh S Dhall A Hayat M Knibbe J Ji Q Automatic gaze analysis: A survey of deep learning based approaches Ieee T Pattern Anal 2024 46 1 61 84 10.1109/TPAMI.2023.3321337 37966935 Ghosh S, Dhall A, Hayat M, Knibbe J, Ji Q. Automatic gaze analysis: A survey of deep learning based approaches. Ieee T Pattern Anal. 2024;46(1):61–84. 10.1109/TPAMI.2023.3321337 37966935
54. Huang L Qin J Zhou Y Zhu F Liu L Shao L Normalization techniques in training dnns: Methodology, analysis and application Ieee T Pattern Anal 2023 45 8 10173 96 10.1109/TPAMI.2023.3250241 37027763 Huang L, Qin J, Zhou Y, Zhu F, Liu L, Shao L. Normalization techniques in training dnns: Methodology, analysis and application. Ieee T Pattern Anal. 2023;45(8):10173–96. 10.1109/TPAMI.2023.3250241 37027763
55. Ahsan MM Mahmud MP Saha PK Gupta KD Siddique Z Effect of data scaling methods on machine learning algorithms and model performance Technologies 2021 9 3 52 10.3390/technologies9030052 Ahsan MM, Mahmud MP, Saha PK, Gupta KD, Siddique Z. Effect of data scaling methods on machine learning algorithms and model performance. Technologies. 2021;9(3):52.
56. Hastie T Tibshirani R Friedman J The Elements of Statistical Learning. 2009 Springer Hastie T, Tibshirani R, Friedman J. The Elements of Statistical Learning. Springer; 2009. 10.1007/978-0-387-84858-7.
57. James G Witten D Hastie T Tibshirani R An Introduction to Statistical Learning: with Applications in R. 2021 Springer James G, Witten D, Hastie T, Tibshirani R. An Introduction to Statistical Learning: with Applications in R. Springer; 2021. 10.1007/978-1-0716-1418-1.
58. Batten B, Hosseini M, Lomuscio A. Tight verification of probabilistic robustness in bayesian neural networks. arXiv preprint arXiv:2401.11627 (2024.
59. Ataş PK. A novel hybrid model to predict concomitant diseases for hashimoto’s thyroiditis. BMC Bioinf. 2023;24(1). 10.1186/s12859-023-05443-5. 10.1186/s12859-023-05443-5 PMC10464155 37620755
60. Shaikh TA, Ali R, Beg MMS. Transfer learning privileged information fuels cad diagnosis of breast cancer. Mach Vis Appl. 2020;31(1–2). 10.1007/s00138-020-01058-5.
61. Karadayi-Ataş P Sevkli AZ Tufan K A vns based framework for early diagnosis of the alzheimer’s disease converted from mild cognitive impairment Optim Lett 2021 17 9 2347 66 10.1007/s11590-021-01816-y Karadayi-Ataş P, Sevkli AZ, Tufan K. A vns based framework for early diagnosis of the alzheimer’s disease converted from mild cognitive impairment. Optim Lett. 2021;17(9):2347–66. 10.1007/s11590-021-01816-y.
62. Ataş PK Evaluate student achievement by classifying brain structure and its functionality with novel hybrid method Neural Comput Appl 2023 36 7 3357 68 10.1007/s00521-023-09031-9 Ataş PK. Evaluate student achievement by classifying brain structure and its functionality with novel hybrid method. Neural Comput Appl. 2023;36(7):3357–68. 10.1007/s00521-023-09031-9.
63. Shaikh TA Ali R Automated atrophy assessment for alzheimer’s disease diagnosis from brain mri images Magnetic Reson Imag 2019 62 167 73 10.1016/j.mri.2019.06.019 31279772 Shaikh TA, Ali R. Automated atrophy assessment for alzheimer’s disease diagnosis from brain mri images. Magnetic Reson Imag. 2019;62:167–73. 10.1016/j.mri.2019.06.019. 10.1016/j.mri.2019.06.019 31279772
64. Karadayı Ataş P:. A novel clustered-based binary grey wolf optimizer to solve the feature selection problem for uncovering the genetic links between non-hodgkin lymphomas and rheumatologic diseases. Health Inf Sci Syst. 2025;13(1). 10.1007/s13755-025-00350-w. 10.1007/s13755-025-00350-w PMC12048384 40321894
65. Shaikh TA Ali R An intelligent healthcare system for optimized breast cancer diagnosis using harmony search and simulated annealing (hs-sa) algorithm Inf Med Unlocked 2020 21 100408 10.1016/j.imu.2020.100408 Shaikh TA, Ali R. An intelligent healthcare system for optimized breast cancer diagnosis using harmony search and simulated annealing (hs-sa) algorithm. Inf Med Unlocked. 2020;21:100408. 10.1016/j.imu.2020.100408.
66. Hassija V Chamola V Mahapatra A Singal A Goel D Huang K Scardapane S Spinelli I Mahmud M Hussain A Interpreting black-box models: a review on explainable artificial intelligence Cognit Comput 2024 16 1 45 74 10.1007/s12559-023-10179-8 Hassija V, Chamola V, Mahapatra A, Singal A, Goel D, Huang K, Scardapane S, Spinelli I, Mahmud M, Hussain A. Interpreting black-box models: a review on explainable artificial intelligence. Cognit Comput. 2024;16(1):45–74.
67. Dwivedi R Dave D Naik H Singhal S Omer R Patel P Qian B Wen Z Shah T Morgan G Explainable ai (xai): Core ideas, techniques, and solutions ACM Comput Surv 2023 55 9 1 33 10.1145/3561048 Dwivedi R, Dave D, Naik H, Singhal S, Omer R, Patel P, Qian B, Wen Z, Shah T, Morgan G, et al. Explainable ai (xai): Core ideas, techniques, and solutions. ACM Comput Surv. 2023;55(9):1–33.
68. Badawy M. Integrating artificial intelligence and big data into smart healthcare systems: A comprehensive review of current practices and future directions. Artif Intell Evolution. 2023;133–53.
69. Hakkoum H Idri A Abnane I Global and local interpretability techniques of supervised machine learning black box models for numerical medical data Eng Appl Artif Intel 2024 131 107829 10.1016/j.engappai.2023.107829 Hakkoum H, Idri A, Abnane I. Global and local interpretability techniques of supervised machine learning black box models for numerical medical data. Eng Appl Artif Intel. 2024;131:107829.
70. Singla K, Biswas S. Machine learning explanability method for the multi-label classification model. 2021 IEEE 15th International Conference on Semantic Computing (ICSC). 2021. 337–40. IEEE.
71. Davagdorj K Li M Ryu KH Advances in Intelligent Information Hiding and Multimedia Signal Processing: Proceeding of the 16th International Conference on IIHMSP in Conjunction with the 13th International Conference on FITAT 2021 Ho Chi Minh City, Vietnam, Springer 426 33 Davagdorj K, Li M, Ryu KH: Local interpretable model-agnostic explanations of predictive models for hypertension. In: Advances in Intelligent Information Hiding and Multimedia Signal Processing: Proceedings of the 16th International Conference on IIHMSP in Conjunction with the 13th International Conference on FITAT; 2020 Nov 5–7; Ho Chi Minh City, Vietnam. Vol. 2. Cham: Springer; 2021. p. 426–33.
72. Sousa I Maria Bernardes Rebuzzi Vellasco M Silva E Local interpretable modelagnostic explanations for classification of lymph node metastases Sensors 2019 19 13 2969 10.3390/s19132969 31284419 PMC6651753 Sousa I, Maria Bernardes Rebuzzi Vellasco M, Silva E. Local interpretable modelagnostic explanations for classification of lymph node metastases. Sensors. 2019;19(13):2969. 31284419 10.3390/s19132969 PMC6651753