Predicting Student Performance in Online Learning Using a Highly Efficient Gradient Boosting Decision Tree

Abstract

Online learning has become a popular way of learning due to the rapid development of informatization of education, the rise of online learning platforms has provided great convenience to students’ learning, breaking the limitations of time and space, enabling students to study various courses anytime, anywhere. However, due to the lack of an effective supervision mechanism, online learning undermined by low quality tutoring and a substantial dropout rate of students. In this study, we used an efficient ensemble learning algorithm, LightGBM, to develop a student performance prediction model. Basing on the interaction behavior data of students’ online learning, the prediction model can predict whether a student will be able to pass the course. As a result of identifying at-risk students, teachers can provide targeted interventions to improve the learning performance of these students. During the experiments, we compared with ten classical machine learning algorithms on a public dataset. It reflected that LightGBM outperformed in predicting student performance. The study also analyzed the impact of different interaction behaviors on students’ online learning performance, which will help teachers and educational researchers to better understand the relationship between students’ online learning behavior and students’ learning performance. This will help teachers guide students in online learning and optimize the design of online learning courses.