A novel optimal feature selection technique for medical data classification using ANOVA based whale optimization

Abstract

This research, a hybrid model to feature selection for classification of different heart disease (HD) dataset is introduced. At first and, a filter method has been utilized to select the relevant feature sets from the actual feature sets, in particular ANOVA. At that point, an evolutionary wrapper-based methodology using whale optimization (WO) to discover the optimal feature sets from the previous feature selection is proposed. The primary target of utilizing WO is to tune into three stages. To start with, whale calculations are used to identify the whole features to dispose of half of the less significant features. Secondary, the congestions of mutual are utilized to make the rest of the features a priority and arrange. Tertiary, WO determines the majority of the 10 best features that use forward features. The support vector machine, K-nearest neighbor, and Naïve Bays have been utilized in the selection of the optimal feature set of coronary disease results. The ANOVA-WO technique is tested with binary and multi-class HD datasets to perform a complete analysis study. Since outcome investigation, better classification accuracy with extensively fewer features than that of the benchmark plans has been obtained through the proposed approaches.