Применение алгоритмов машинного обучения для прогнозирования академической успеваемости курсантов

Abstract

The modern development of military education requires transitioning from reactive to proactive management of aca demic performance through early identification of cadets at risk of academic difficulties. This is particularly important in military institutions, where cadets face heavy workloads and limited time for remediation, while traditional assessment systems often reveal problems too late for effective intervention. The substantial volume of digital data generated during training—electronic grade books, test results, and activity logs—can serve as a source of features for predicting future performance. Machine learning methods, partic ularly ensemble algorithms such as gradient boosting over decision trees (e.g., XGBoost), are highly effective for this task due to their ability to handle nonlinear dependencies and heterogeneous features while maintaining interpretability for educational decision making.