Artificial Intelligence-Based Prediction of Student Academic Performance: An Explainable Machine Learning Approach to AI-Assisted Learning

Abstract

Artificial intelligence-assisted learning is increasingly influencing higher education, creating a need for predictive approaches that can identify academic-performance patterns while remaining interpretable. This study aimed to develop and explain machine-learning models for predicting student academic performance using demographic and learning-related characteristics. Secondary data from 257 students were analyzed, with CGPA as the outcome and age, gender, academic level, location, Creative Thinking, GenAI Literacy, Digital Literacy, and GenAI use frequency as predictors. Elastic Net, Random Forest, Gradient Boosting, and support vector regression were compared using five-fold cross-validation and a held-out 20% test set. Random Forest achieved the lowest mean cross-validation RMSE (0.489) and was retained as the final model. On the test set, it achieved R² = 0.382, RMSE = 0.420, and MAE = 0.299. Creative Thinking and Age emerged as the most influential predictors, while GenAI Literacy also contributed meaningfully to prediction. SHAP analysis identified positive contributions for Creative Thinking, a nonlinear pattern for Age, and an inverse conditional pattern for GenAI Literacy. The findings demonstrate that explainable machine learning can provide moderate predictive performance while revealing distinct predictor contributions, supporting transparent academic-performance analytics in AI-assisted learning environments.