Optimizing Predictive Accuracy Through Ranking-Based Filter Feature Selection

Abstract

Irrelevant information within datasets can hinder both the speed and accuracy of classification models. This study aims to identify the optimal feature set for building a more efficient model. We investigated the impact of using both the full and selected feature sets of Chronic Kidney Disease (CKD) data on the performance of classification models in supervised machine learning. Our methodology involved utilizing the Adaptive Neuro-Fuzzy Inference System (ANFIS) classification model, applying feature selection to isolate the most relevant predictors for CKD. The findings highlight the significance of feature selection, with the selected features showing improved performance—achieving sensitivity of 100%, specificity of 76.92%, and accuracy of 87.5%—compared to using the full feature set. This research underscores the importance of feature selection in enhancing classification model effectiveness, particularly for medical applications.