DNFH: Data Normalization And Feature Selection Using Hybrid Methods For Heart Disease Prediction

Abstract

Death from heart disease has consistently ranked high among the leading causes of mortality in the globe. Predicting the likelihood of getting heart disease based on certain features is crucial because of the high cost of identifying heart disease. Feature extraction is an important step in any classification system as it reduces dimensionality and enhances accuracy. This paper proposes a technique called Data Normalization and Feature Selection Using Hybrid Methods (DNFH) for predicting the likelihood of getting heart disease based on selected features. This technique uses Weighted Transform K-Means Clustering (WTKMC) for data normalization; Elastic Net (EN), Random Forest Classifier (RFC), Binary BAT (BBA), and Weighted Binary BAT Algorithm (WBBAT) for ensemble feature selection. Experiments were conducted using the Cleveland dataset and classification accuracy was compared with Support Vector Machine (SVM) and Logistic Regression (LR) models. The results showed significant improvement in accuracy using WTKMC for preprocessing and ensemble feature selection techniques.