Prediction of Academic Performance of Student Cadets at the Republic of Indonesia Defense University Using Random Forest

Abstract

The academic readiness of cadets at the Indonesian Defense University (UNHAN RI) needs proactive monitoring, as cadets carry both academic workloads and military discipline simultaneously. This study aims to build a classification model for the academic performance of UNHAN RI cadets into three categories, Low, Moderate, and High, using the Random Forest algorithm within the CRISP-DM framework. Data were collected through an online questionnaire distributed to 100 cadet respondents from cohorts 5 and 6. After data cleaning, 95 valid records were obtained. The academic performance label was constructed from academic grades using a mean ± standard deviation categorization method, yielding a distribution of 20.00% Low, 57.89% Moderate, and 22.11% High. Six daily-habit variables, namely study hours, assignment load, sleep hours, exercise frequency, academic fatigue level, and perceived task burden, were used as predictors without including the grade itself, to avoid data leakage. Data were split into 80% training and 20% test sets using stratified sampling. A Random Forest model with 100 decision trees achieved 57.89% accuracy on the test set and an average of 55.79% across 5-fold cross-validation. Study hours, exercise frequency, and assignment load were identified as the three most influential features. An experiment removing the least important feature, perceived task burden, produced comparable accuracy and cross-validation mean but a higher cross-validation standard deviation, so the original six features were retained in the final model. The model's accuracy falls into the Poor category under common interpretation guidelines and is close to the majority-class baseline, indicating that daily-habit data alone is not yet sufficient to reliably predict academic performance at this limited sample size. The model is intended as a starting point for the institution to develop a more mature early-detection system for cadets' academic performance.