ETCD: An effective machine learning based technique for cardiac disease prediction with optimal feature subset selection

Abstract

Cardiac disease is the leading cause of death worldwide. The early diagnosis and prognosis can help patients live longer by lowering mortality and boosting survival rates. The paucity of radiologists and doctors in various nations, due to a variety of factors, is a substantial barrier to early diagnosis. Computational intelligence is an emerging concept in the field of medical imaging to identify, prognosticate, and diagnose disease, among numerous initiatives to construct decision support systems. It relieves radiologists and doctors from being overworked and reduces the time it takes to diagnose patients promptly. In this work, an effective technique for cardiac disease (ETCD) prediction based on machine intelligence has been proposed. To ensure the success of our proposed model, we used effective Data Collection, Data Pre-processing, and feature selection process to generate accurate data for the training model. ETCD utilizes the optimal feature subset selection algorithm (OFSSA) to extract features from different datasets (Cleveland, Hungarian, Combined dataset, and Z_Alizadeh Saini datasets) having varying properties available at the UCI machine learning repository. With ETCD, the average accuracy performance for considered datasets gets increased with SVM, KNN, DT, NB, and RF classifiers by 6.227%, 2.72%, 7.345%, 14.084%, and 18.921% respectively. Further, the results of the experiments demonstrate that ETCD outperformed several contemporary baseline approaches in terms of accuracy and was comparable in terms of sensitivity, specificity, precision, and F_Score. ETCD returns the best feasible solution among all input predictive models considering performance criteria and improves the efficacy of the system, hence can assist doctors and radiologists in a better way to diagnose cardiac patients.