Predicting Academic Achievement in Artificial Intelligence Education Using Student Assessment Indicators

Abstract

In Higher Education, prediction of student performance based on assessment signals has become a critical task in accurately evaluating performance and as a support to data-driven decision making. This study aimed to evaluate the final exam grades prediction performance of machine learning algorithms for the Artificial Intelligence Course students. Data were 560 student records and data were analyzed using quantitative predictive approach. Three predictive indicators – Quiz scores, Midterm examination marks and three assignment scores – were analyzed. Four regression models namely Linear Regression, Random Forest, Support Vector Regression and XGBoost were tested based on mean absolute error (MAE), root mean square error (RMSE) and coefficient of determination (R²). The model with the highest predictive accuracy was XGBoost with an RMSE of 5.252, an MAE of 3.949 and an R² of 0.393. The feature importance analysis showed that the important predictors were the performance of the midterm exam and Assignment 3. The results do indicate the potential for using assessment-based machine learning models for academic monitoring and educational decision making; however, the accuracy was moderate. A model should be further tested by other education contexts to enhance the generalizability of the model.