Harnessing Ridge Regression and SHAP for Predicting Student Grades: An Approach Towards Explainable AI in Education

This paper presents a comparative analysis of several regression techniques for predicting student academic performance, with a particular focus on Ridge Regression and its interpretability through SHAP, a method of Explainable Artificial Intelligence (XAI). The models evaluated include Linear Regression, Ridge Regression, Lasso Regression, Elastic Net, Decision Tree Regression, Random Forest Regression, AdaBoost, Gradient Boosting, Bagging, XGBoost, and K-Nearest Neighbors. Using a dataset encompassing student family background, personal information, and recent academic grades, we trained and evaluated these models to predict future academic performance. Experimental results demonstrated that Ridge Regression outperformed all other models in predictive accuracy with the highest r2 score of 0.9036 and lowest RMSE of 1.4623. Furthermore, we employed SHAP values to interpret the predictions made by the Ridge Regression model, revealing key feature contributions to the model’s output. This research underscores the potential of Ridge Regression and SHAP in building predictive and interpretable models in the educational domain. This fusion of predictive accuracy and interpretability provides invaluable insights for educators and policy-makers alike in understanding and enhancing student academic achievement.