Experimental Analysis of Filtering Based Feature Selection Techniques for Fetal Health Classification

Abstract

Machine learning techniques enable machines to acquire intelligence through learning. Trained machines carry out various tasks including prediction, classification, clustering and recommendation that are widely used in dispersed applications. Classification, a supervised learning technique can be improved using feature selection techniques viz., filtering based, wrapper based and embedded based. This paper explores the impact of filtering based feature selection techniques on classification technique. In correlation based filtering technique, Pearson correlation, Spearman correlation and Kendall rank correlation has been identified for analysis. Similarly, in statistical filtering technique, mutual information, chi squared score, ANOVA univariate test and Univariate ROC-AUC based techniques are analysed. These filtering techniques are evaluated on K nearest neighbor, Support vector machine, decision trees and Gaussian naïve bayes classification techniques. The experimental analyses were carried out using fetal heart rate dataset. The performance analyses were carried out in terms of precision, recall, f1 score and accuracy.