<![CDATA[Academic performance prediction is a crucial area in education; however, the complexity of influencing factors often cannot be adequately captured by simple linear models. This research conducts a methodological comparative analysis of five machine learning models Simple Linear Regression, Multiple Linear Regression (MLR), Decision Tree, Random Forest, and Artificial Neural Network (ANN) to determine the most accurate predictive approach using a comprehensive dataset encompassing academic, behavioral, and psychosocial factors. The models were evaluated using Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and R-squared (R2) metrics. Evaluation results on the test data revealed that the Multiple Linear Regression (MLR) model unexpectedly delivered the most superior performance, achieving an R2 value of 0.7324 and the lowest RMSE of 2.0391. Further analysis from non-linear models identified Attendance and Hours_Studied as the two factors with the highest predictive influence. This study concludes that interpretable models like MLR can be highly effective when supported by relevant features, offering practical implications for institutions to develop effective early warning systems by focusing on key, actionable factors.]]>
Copyrights © 2025
