Privacy-aware vaccine recommendation using federated learning and blockchain.

Suitable vaccines for individuals are suggested by the vaccine recommendation system regarding certain criteria. Nevertheless, the existing studies didn't augment the vaccine recommendation system centered on users' symptoms and medical history among several geographical locations in the hyperledger fabric blockchain. Thus, in this paper, Krichevsky Dirichlet Trofimov-based latent Dirichlet allocation (KDT-LDA) and federated learning-Expcos bidirectional distillation long short-term memory (FL-EBiDLSTM)-based vaccine recommendation systems using symptoms and medical history are presented. Primarily, the vaccine symptoms dataset is taken. Then, the pre-processing is done based on named entity recognition, tokenization, and stemming. Later, by employing NSR-KMeans, the pre-processed data is grouped. Later, KDT-LDA-based symptoms and medical history modeling and adversarial debiasing-ClinicalBERT-based word embedding are carried out. Simultaneously, from the pre-processed data, the polarity score is identified. By utilizing the Cauchy Cubic-based fuzzy inference system, the labelling is performed regarding the polarity score. After that, the labelling outcomes are trained by the EBiDLSTM-based sentiment nature identification. Then, the natural language processing features are extracted from the symptoms and medical history modeling outcomes. After that, by using Spearman rank correlation, feature correlation is performed. Lastly, the appropriate vaccine is predicted based on FL-EBiDLSTM. Here, to solve the issue of training the patient data among various locations, FL is included. In real-time, vaccine demand users register with the hyperledger fabric blockchain and upload their medical history. Later, the vaccine recommendation system suggests the vaccines concerning the history. As per the outcomes, the proposed model achieved a high accuracy of 99% and outperformed prevailing techniques.