Performance Evaluation of Embedded Feature Selection Techniques over Filter and Wrapper-based Feature Selection Techniques

This paper compared the performance of filter-based and wrapper-based feature selection techniques with that of embedded-based feature selection technique. Recursive Feature Elimination (RFE) with SVM Feature selection techniques were explored on Random Forest (F) and Logistic Regression (RFE-LR). Recursive Feature Elimination (RFE) with SVM Feature selection techniques were explored on Random Forest (RF) and Logistic Regression (LR) using ten (10) publicly available defect datasets. SVM-RFE-LR approach of Embedded Feature selection Techniques produced consistent software metrics within the range o This paper compared the performance of filter-based and wrapper-based feature selection techniques with that of embedded-based feature selection technique. Recursive Feature Elimination (RFE) with SVM Feature selection techniques were explored on Random Forest (F) and Logistic Regression (RFE-LR). Recursive Feature Elimination (RFE) with SVM Feature selection techniques were explored on Random Forest (RF) and Logistic Regression (LR) using ten (10) publicly available defect datasets. SVM-RFE-LR approach of Embedded Feature selection Techniques produced consistent software metrics within the range of 18% to 55% across the datasets, and SVM-RFE-RF reported within 33% to 96%. Significance performance was witnessed over Filter and Wrapper based techniques. f 18% to 55% across the datasets, and SVM-RFE-RF reported within 33% to 96%. Significance performance was witnessed over Filter and Wrapper based techniques.