Monitoring student learning outcomes at educational institutions is a significant concern for researchers and parents. Determining student learning outcomes is challenging because student learning outcomes depend on different factors. By leveraging some classical machine learning algorithms, we evaluate the prediction results on student academic performance with various factors. Finding important factors helps improve the prediction model’s accuracy, providing additional insights into the relationships between factors. The study demonstrates that the performance of classical machine learning algorithms can be slightly improved with the GridSearchCV and Optuna methods, with Optuna performing better than all. In addition, the performance on about four factors can reach 0.7 in the F1-Score compared to around 0.77 on the original features, reassuring the effectiveness of our approach.