Implementasi Algoritma Random Forest untuk Prediksi Waktu Penyelesaian Hafalan Al-Qur’an Berbasis Website

Abstract

Manual monitoring of Quranic memorization (tahfizh) in Islamic boarding schools faces efficiency challenges due to large student populations and paper-based record keeping. This study aims to implement the Random Forest algorithm to predict the estimated completion time of Quranic memorization in a web-based monitoring system at Madrasah Aliyah Jam’iyyah Islamiyyah, Tangerang Selatan, Indonesia. The dataset consists of 12,458 memorization logs from 271 students during March 1 to May 3, 2026. Feature engineering produced 15 features covering Quranic text complexity, student memorization history, and temporal patterns; Spearman correlation feature selection reduced these to 13 significant features. The model was optimized using GridSearchCV and evaluated with MAE, RMSE, R², MAPE, and 5-fold cross-validation. Random Forest achieves R²=0.8966, MAE=0.6141, and MAPE=6.98% on the 70:30 split, outperforming Decision Tree (R²=0.8879) and matching XGBoost (R²=0.8964). Cross-validation yields CV R²=0.9004, confirming stable generalization. Feature importance analysis indicates that student learning habits are stronger predictors than Quranic text complexity. As a practical contribution, the model is integrated into a web-based monitoring system enabling teachers to track all students’ progress centrally and receive automated memorization completion estimates, enhancing the effectiveness of guidance in tahfizh institutions.