Experimental analysis of filtering-based feature selection techniques for fetal health classification

Machine learning techniques enable computers to acquire intelligence through learning. Trained machines can carry out various tasks, such as prediction, classification, clustering, and recommendation, within a wide variety of applications. Classification is supervised technique that be improved using feature selection filtering, wrapping, embedding. This paper explores the impact filtering-based on classification methods, focuses an analysis correlationbased filtering based Pearson, Spearman, Kendall rank correlation. Similarly, we explore impacts statistical mutual information, chi-squared score, ANOVA univariate test, ROC-AUC. These are evaluated by implementing them with k-nearest neighbor, support vector machine, decision tree, Gaussian na?ve Bayes methods. Our experiments were carried fetal heart rate dataset, performance each combination methods was measured precision, recall, F1-score, accuracy. An experimental results showed metrics for neighbor 3% use technique, 4% improvement observed tree machine correlation-based technique. Of techniques, ROC-AUC best they accuracy 92%; compared other correlation Spearman coefficient gave results, it also 92%.