Abstrak Program Makan Bergizi Gratis (MBG) bertujuan meningkatkan gizi siswa dan menekan stunting. Pengawasan manual komposisi gizi menu rentan human error dan tidak efisien untuk data besar. Penelitian ini membangun model klasifikasi kelayakan menu berbasis Decision Tree dengan komparasi Naïve Bayes. Data berupa 5.000 record dataset nutrisi Kaggle meliputi protein, karbohidrat, lemak, serat, gula, kalori, dan label calorie_class. Tahapan meliputi preprocessing, eliminasi fitur kalori dengan operator Select Attributes untuk mencegah data leakage, pemodelan paralel, serta evaluasi menggunakan skema 10-Fold Cross Validation dan Confusion Matrix. Hasil penelitian menunjukkan setelah kalori diisolasi, parameter karbohidrat bergeser menjadi root node pada Decision Tree. Model Decision Tree mencatatkan akurasi riil sebesar 87,82%, sedangkan Naïve Bayes menghasilkan 86,46%. Evaluasi metrik menunjukkan kedua model memiliki performa kuat, stabil, dan seimbang dalam memetakan seluruh kelas target. Temuan ini membuktikan bahwa pendekatan komparatif data mining efektif untuk mengklasifikasikan kelayakan menu MBG secara objektif, transparan, dan reliabel sebagai sistem pendukung keputusan. Kata Kunci: Decision Tree, Naïve Bayes, Data Mining, Klasifikasi Gizi, Makan Bergizi Gratis. Abstract The Free Nutritious Meal (MBG) program aims to improve students' nutritional quality and reduce stunting. Manual monitoring of menu alignment remains vulnerable to human error and inefficient for large datasets. This study develops a Decision Tree classification model compared with the Naïve Bayes algorithm to support MBG menu evaluation. The dataset comprises 5,000 Kaggle nutrition records containing protein, carbohydrate, fat, fiber, sugar, calorie, and calorie_class variables. The stages involved preprocessing, eliminating the calorie feature via the Select Attributes operator to prevent data leakage, parallel modeling, and evaluation using a 10-Fold Cross Validation scheme and a Confusion Matrix.The results show that after isolating the calorie attribute, the carbohydrate parameter (carbs_g) shifted to the root node in the Decision Tree. The Decision Tree model achieved a real accuracy of 87.82%, while Naïve Bayes yielded 86.46%. Metric evaluations indicate that both models possess robust, stable, and balanced performance across all target classes. These findings prove that a comparative data mining approach is effective for classifying MBG menu feasibility objectively, transparently, and reliably as a decision support system. Keywords: Decision Tree, Naïve Bayes, Data Mining, Nutritional Classification, Free Nutritious Meal.
Copyrights © 2026