Explainable AI for Early Prediction of Student Academic Success Using Learning Behaviour Analytics

Abstract

The increasing availability of digital learning platforms has created opportunities to identify students who may experience academic difficulties before poor outcomes become evident. Learning management systems continuously generate behavioural traces, including login activity, assessment participation, assignment submission patterns, resource engagement, and interaction with course materials. When systematically analysed, these behavioural indicators can provide valuable signals for the early prediction of student academic success. However, conventional machine-learning approaches often provide predictions without sufficiently explaining the factors responsible for those predictions, limiting their usefulness and acceptance among educators. This study proposes an explainable artificial intelligence framework for the early prediction of student academic success using learning behaviour analytics. The framework integrates behavioural data preprocessing, temporal feature engineering, machine-learning classification, and explainable AI techniques to generate both outcome predictions and interpretable explanations. Multiple predictive models are considered to examine their ability to distinguish students according to their expected academic outcomes at progressively earlier stages of the learning period. Explainability techniques are subsequently employed to identify the behavioural characteristics that contribute most strongly to model predictions at both global and individual levels. The proposed approach is intended to move beyond prediction alone by providing educators with actionable information about the behavioural patterns associated with academic risk or success. By examining prediction performance across different points in the learning process, the study also addresses the practical question of how early reliable intervention signals can be generated. The resulting framework provides a transparent foundation for data-informed academic support while highlighting the importance of interpretability, responsible use of student data, and timely educational intervention.