Explaınable Machıne Learnıng For Predıctıng Students’ Academıc Performance From Generatıve AI Usage Patterns in Hıgher Educatıon

Generative artificial intelligence is increasingly embedded in higher education learning practices, yet its contribution to academic performance remains contested. This study developed an explainable machine-learning framework for predicting students’ post-semester GPA from Generative AI usage patterns and learning-related characteristics. Using a structured dataset of 50,000 student records, the study modelled academic performance with predictors including pre-semester GPA, weekly GenAI usage hours, primary use case, prompt-engineering skill, tool diversity, traditional study hours, perceived AI dependency, institutional policy, exam anxiety, year of study, and major category. To prevent outcome leakage, non-relevant outcome variables and student identifiers were excluded before training. Ridge Regression, Decision Tree Regression, and Extra Trees Regression were compared using R², RMSE, and MAE. The results showed that Ridge Regression achieved the strongest predictive performance, indicating that academic outcomes were largely explainable through stable relationships among prior achievement, GenAI-related behaviours, and study characteristics. Feature-importance analysis revealed that prior GPA, primary GenAI use case, and prompt-engineering skill were among the most informative predictors. The study contributes to GenAI-aware learning analytics by demonstrating how explainable modelling can support evidence-based academic monitoring.