Predicting the Academic Performance of Programming Students Using Logs from an Online Learning System: Toward Providing Timely Guidance and Feedback to Students

In this study, we developed a programming learning system that presents instructors with the trends of an entire class and the coding status of individual students in programming practice classes. A survey filled by faculty members confirmed that the “understand students’ coding status” objective was sufficiently satisfied, and that the system helped faculty members provide appropriate guidance and timely advice to students. In addition, we created and evaluated a multiple regression analysis model using the operation logs from the programming exercise system for predictions and effective feedback provision to students. With the model, we could predict final exam scores using system operation logs from the early-stages of the course. Additionally, we could update the score predictions based on each week’s class logs and provide feedback to students. This predictive model has implications for educational practice and allows for timely interventions for learner success. Furthermore, we confirmed the model’s good performance by comparing the scores predicted using past data with the actual scores. The study is particularly strong in identifying key variables and presenting detailed statistical analyses. This makes its conclusions both valuable and actionable for educators. We believe that this system be instrumental in improving the performance of students in programming classes.