A Machine Learning Framework for Heart Disease Prediction Using Swarm Intelligence

Abstract

Cardiovascular disease continues to be one of the main causes of death around the globe, highlighting the importance of developing precise and dependable predictive models which can facilitate the diagnosis at an early stage and help in making clinical decisions. The last few years have seen the usage of machine learning and deep learning techniques to predict heart disease from structured clinical data with a strong potential; however, obtaining consistently high predictive performance along with model robustness has remained a challenge. The paper presents a new framework for heart disease prediction that is based on the swarm intelligence and ensemble learning concepts and uses tabular clinical data. The method we suggest couples the particle swarm optimization-based feature selection with a swarm-optimized ensemble weighting strategy which allows for the adaptive combination of several complementary classifiers, such as the gradient-boosted decision trees, randomized tree ensembles, support vector machines, and logistic regression models. The robust framework will automatically find the informative feature subsets and will also adjust the ensemble contributions for the maximum predictive accuracy. The proposed model was put to the test on a prominent heart disease dataset that contains 1,025 clinical records along with 13 input features. The results obtained from the experiments indicate that the newly developed swarm-optimized ensemble has 100% classification accuracy, AUC of 1.00, sensitivity of 100%, and specificity of 100%, which are all the same or better than the results reported for the best machine learning and deep learning techniques in the literature. Furthermore, the suggested method is still choosing small feature subsets consistently while getting almost no decrease in predictive performance. This means that the use of swarm intelligence as an optimizer is very effective when it comes to the area of predictive modeling of heart disease with an ensemble approach