Optimized heart disease prediction model using a meta-heuristic feature selection with improved binary salp swarm algorithm and stacking classifier.

Despite technological advancements, heart disease continues to be a major global health challenge, emphasizing the importance of developing accurate predictive models for early detection and timely intervention. This study proposes a heart disease prediction model integrating a stacking classifier with a nature-inspired meta-heuristic algorithm. It employs an improved Binary Salp Swarm Algorithm (BSSA) by incorporating a wolf optimizer and opposition-based learning for optimal feature selection. The proposed Stacking Classifier (SC) architecture features a two-tier ensemble: heterogeneous base classifiers at level 0 and a meta-learner at level 1. The BSSA is used to identify optimal features, which are then utilized to construct the stacking classifier. Experimental results demonstrate superior performance, achieving 95 % accuracy, 0.92 sensitivity, 0.97 specificity, 0.96 precision, and an F1 score of 0.95, with notably low false positive and false negative rates. Further, validation on larger datasets yielded an accuracy of 87.46 %. The feature selection process adopts a multi-objective strategy which enhances the classification accuracy and outperforms conventional techniques. The proposed method demonstrates significant potential for improving the predictive modelling in clinical settings for diagnosing heart diseases.