Sistem Rekomendasi Pola Makan Sehat Berdasarkan Prediksi Obesitas

  • Guna Yanti Kemala Sari Siregar Universitas Muhamamdiyah Metro
  • Ika Arthalia Wulandari Universitas Muhammadiyah Metro
Keywords: Obesity, Machine Learning, Gradient Boosting, Random Forest, Recommendation System, Healthy Dietary Pattern

Abstract

Obesity is a health problem associated with dietary patterns and individual physical conditions, creating a need for data-driven approaches to support obesity classification and provide appropriate dietary recommendations. This study aims to develop a healthy dietary recommendation system based on obesity class prediction using machine learning methods. The dataset consists of 2,111 observations with 16 predictor variables. After removing duplicate records, 2,087 observations were retained for analysis. Obesity prediction was performed using Random Forest and Gradient Boosting with an 80:20 training-testing split. Model performance was evaluated using accuracy, precision, recall, F1-score, and stratified five-fold cross-validation. The results show that Gradient Boosting achieved better performance, with an accuracy of 98.56% and a weighted F1-score of 98.57%, while the mean accuracy obtained through cross-validation was 96.74% ± 1.15%. Feature importance analysis indicated that Weight, Height, FCVC, and Gender were among the most influential characteristics for prediction. The prediction results were subsequently combined with users' behavioral characteristics through a rule-based recommendation engine to generate more targeted healthy dietary recommendations. The recommendation engine triggered at least one rule for 95.54% of the analyzed data. These findings indicate that integrating obesity prediction with a rule-based recommendation engine can provide a data-driven approach to supporting healthy dietary recommendations based on user characteristics.

Published
2026-10-28
Section
Articles