<![CDATA[The problem of student dropout is a strategic issue in higher education because it has a direct impact on academic quality, institutional efficiency, and university accreditation. This study aims to statistically analyze the factors that contribute to the variation in GPA of students who have dropped out using the Explainable Machine Learning approach. The predictive model was built using the Extreme Gradient Boosting (XGBoost) algorithm to obtain optimal prediction performance, while the Shapley Additive Explanations (SHAP) method was used to provide interpretation of the contribution of each feature in the model. The research dataset includes academic, administrative, and demographic data of students who have dropped out in the last two academic years. The evaluation results show that the XGBoost model shows excellent predictive performance with an R² value of 0.820 indicating that most of the GPA variation can be explained by the model, and is supported by a Root Mean Squared Error (RMSE) value of 0.344 and a Mean Absolute Error (MAE) of 0.172 indicating that the prediction error rate is relatively low. SHAP analysis revealed that the number of credits taken and tuition payment status were the two factors that statistically significantly contributed to GPA changes in the predictive model. This study provides more comprehensive insights by combining high predictive performance and model interpretability, enabling educational institutions to identify student academic risk earlier and based on data.]]>
Copyrights © 2026
