Comparison of Forward Selection and Backward Elimination Feature Selection Methods in Support Vector Machine Algorithm

Abstract

Support Vector Machine (SVM) is an effective and robust classification method, particularly when applied to high-dimensional data. However, high-dimensional data often contain irrelevant features that can lead to suboptimal SVM performance. Therefore, a feature selection process is necessary to optimize classification performance by eliminating irrelevant and redundant features from the original dataset. This research aims to compare the Forward Selection and Backward Elimination feature selection methods within the Support Vector Machine Algorithm for classification using the Poverty Depth Index data in Papua Province. The results indicated that applying the Support Vector Machine with Forward Selection feature selection achieved a classification accuracy of 93%, whereas Backward Elimination feature selection achieved a classification accuracy of 97%. Based on these classification accuracy results, it can be concluded that applying Support Vector Machine with Backward Elimination feature selection results in better performance than Forward Selection.