A Comparison of Feature Selection Approaches for Liver Disease Data
Ayushi Pillay, Monika Arya, Sumit Kumar Sar, Bhupesh Kumar Dewangan
Abstract
Efficient feature selection (FS) methods are often needed to correctly classify multiple types of diseases, recognize disease symptoms, and enhance treatment modalities. The aim of this study is to compare the filter and wrapper FS approaches and to show how they enhance classification. First, on disease data, the FS approach has been employed with different classification models. We compared the effectiveness of two FS techniques: recursive feature elimination (RFE) and chi-square feature (CSF) selection. Logistic Regression (LR), Support Vector Machine (SVM), and Decision Tree are three different models (DT) that are used. Performance in classification is evaluated using K-fold crossvalidation. The methods are evaluated on datasets of diseases that are openly accessible. The findings showed that FS is important for accurate disease classification. A subset of the most prevalent characteristics is obtained after examining the relationship between particular symptoms and disease. Thus, by using FS that selects only a few disease predictors, the classification accuracy can be significantly improved.