Perbandingan Random Forest dan K-Nearest Neighbors untuk Klasifikasi Body Mass Index Menggunakan SMOTE-ENN untuk Mengatasi Ketidakseimbangan Data pada Analisis Kesehatan

Naufal Yogi Aptana
Ikhsan, Ali Nur
Maulana Baihaqi, Wiga
Ajeng Widiawati, Chyntia Raras

Infotekmesin
Vol 16 No 1 (2025): Infotekmesin: Januari 2025

This study aims to compare the Random Forest and K-Nearest Neighbors (KNN) algorithms in Body Mass Index (BMI) classification using the SMOTE-ENN method to address data imbalance. The dataset consists of 2111 entries with demographic and health attributes of individuals. Data imbalance poses a significant challenge that may affect the accuracy of machine learning models. The SMOTE-ENN combination was employed to improve data distribution, enabling models to recognize patterns in minority classes better. Key evaluation factors included both algorithms' accuracy, precision, recall, and F1-score. Results indicate that the Random Forest algorithm achieved higher performance with 100% accuracy than KNN with 96% after applying SMOTE-ENN. These findings highlight the unique contribution of SMOTE-ENN in handling imbalanced data, enhancing classification model quality, and significantly impacting machine learning applications in healthcare.