Performance Comparison of SMOTE-Based Ran-dom Forest and Decision Tree Models for Chronic Kidney Disease Classification

Abstract

Chronic Kidney Disease (CKD) is a chronic condition characterized by a gradual decline in kidney function and is often detected at an advanced stage. Previous studies have employed machine learning methods-such as Random Forest and Decision Trees-for CKD classification; however, there is no consensus regarding which algorithm is most optimal for CKD datasets. The large number of clinical variables in CKD data increases computational complexity, necessitating high-performance machine learning methods to improve classification accuracy and support early disease detection. This study aims to develop a CKD classification model and compare the performance of Decision Tree and Random Forest algorithms. The research utilized a health dataset obtained from Mendeley Data. The methodology involved data acquisition, exploratory data analysis, data preprocessing (including addressing class imbalance via the Synthetic Minority Over-sampling Technique, or SMOTE), and model development using Random Forest and Decision Trees, with evaluation performed via 10-fold stratified cross-validation. Model performance was assessed using confusion matrices and classification reports, covering metrics such as accuracy, precision, recall, and F1-score. Experimental results indicate that Random Forest yielded the best classification performance, achieving an accuracy of 0.9437. It is concluded that comparing Random Forest and Decision Tree methods-combined with SMOTE-is an effective approach for identifying the optimal algorithm to enhance accuracy and support decision-making systems for the early detection of CKD.