Implementasi dan Perbandingan Algoritma Random Forest dan Decision Tree untuk Prediksi Risiko Stunting Pada Anak Berbasis Web

ABSTRAK
Stunting merupakan masalah gizi kronis pada balita yang memerlukan deteksi dini.
Penelitian ini membandingkan kinerja algoritma Decision Tree dan Random Forest
untuk klasifikasi status stunting serta mengintegrasikannya ke dalam sistem
berbasis web menggunakan Laravel dan Flask API. Menggunakan 347 data rekam
medis dan penanganan imbalanced data dengan metode Synthetic Minority Over
sampling Technique (SMOTE) pada data latih (80%), hasil evaluasi menunjukkan
bahwa Random Forest memiliki performa superior dengan akurasi 0,86,
dibandingkan Decision Tree yang menghasilkan akurasi 0,78. pada kelas stunting
(Class 1), Random Forest memperoleh precision sebesar 0,36, recall sebesar 0,83,
dan F1-score sebesar 0,50, sedangkan Decision Tree memperoleh precision sebesar
0,24, recall sebesar 0,67, dan F1-score sebesar 0,35. Pada kelas normal (Class 0),
Random Forest memperoleh precision sebesar 0,98, recall sebesar 0,86, dan F1
score sebesar 0,92, sedangkan Decision Tree memperoleh precision sebesar 0,96,
recall sebesar 0,79, dan F1-score sebesar 0,87. Analisis feature importance
mengidentifikasi usia (38,1%) dan tinggi badan (29,6%) sebagai variabel paling
dominan. Model Random Forest berhasil diimplementasikan melalui Flask REST
API dan antarmuka Laravel untuk menyediakan prediksi stunting secara real-time.
Kata Kunci: Stunting, Decision Tree, Random Forest, SMOTE, Laravel, Flask API,
Klasifikasi.
vii
ABSTRACT
Stunting is a chronic nutritional issue among children under five that requires early
detection. This study compares the performance of Decision Tree and Random
Forest algorithms for classifying stunting status and integrates them into a web
based system using Laravel and a Flask API. Utilizing 347 medical records and
addressing data imbalance via the Synthetic Minority Over-sampling Technique
(SMOTE) on the training set (80%), evaluation results demonstrate that Random
Forest outperformed Decision Tree, achieving an accuracy of 0.86 compared to
0.78. For the stunting class (Class 1), Random Forest achieved a precision of 0.36,
recall of 0.83, and F1-score of 0.50, whereas Decision Tree achieved a precision
of 0.24, recall of 0.67, and F1-score of 0.35. For the normal class (Class 0),
Random Forest achieved a precision of 0.98, recall of 0.86, and F1-score of 0.92,
while Decision Tree achieved a precision of 0.96, recall of 0.79, and F1-score of
0.87. Feature importance analysis identified age (38.1%) and height (29.6%) as the
most dominant variables. The Random Forest model was successfully implemented
using a Flask REST API and a Laravel interface to provide real-time stunting
predictions.
Keywords: Stunting, Decision Tree, Random Forest, SMOTE, Laravel, Flask API,
Classification.