A hybrid intelligent model for prediction of coronary artery diseases using TabNet and multiclass SVM

Abstract

Cardiovascular disease is one of the significant fatality-causing diseases in this era by affecting the heart and blood vessels. Cardio diseases are classified into coronary heart disease (CHD), heart failure, valve disease, and arrhythmias. Medical diagnosis of heart disease and treating the patient is a challenging process, where early detection can lead to decreased fatality. In this research, hybrid model-based prediction of CHD detection is developed by TabNet and multiclass support vector machine (SVM). We created our datasets for experimentation by visiting the hospitals in the Mysore and Mandya regions of Karnataka, India. Datasets consist of 16 features; the features are pre-processed to normalize, encode, and handle missing values to extract the aggregate features using TabNet, and the multiclass SVM model is trained to classify the disease based on the classes. The proposed hybrid model prediction performance was evaluated using various metrics such as accuracy, recall, precision, and F1-score.