Predicting Student Achievement before Final Exam: A Regression-Based Approach

An early prediction of student achievement on a course is an important instrument to forecast student accomplishment in the learning process. Besides, it is also functioned to help the students who might be failed. In this study, we proposed a regression-based early prediction of student achievement. The prediction system utilized the data from last semester to predict the final score of the students in the current semester. The proposed approach was tested in two scenarios. First, predicting the final score directly and the second is predicting the final exam score continued by final score calculation by using predefined formula. The experiment result shows that overall, the second scenario produces a better result than the first scenario. The SVR-Linear achieves the highest performance in the first scenario with 3.56 MAE, 0.048 MAPE, 17.63 MSE, 4.2 RMSE, and 0.781 R-squared. SVR-RBF achieves the highest performance in the second scenario with 3.0 MAE, 0.043 MAPE, 13.05 MSE, 3.6 RMSE, and 0.838 R-squared.