MOOC Dropout Prediction Using Learning Process Model and LightGBM Algorithm

With the development and widespread application of Massive Open Online Courses (MOOC) platforms, the issue of reducing learners’ dropout has become a challenge. Therefore, it is necessary to accurately predict whether a learner will complete the course or not. Existing researches mainly use machine learning methods to predict MOOC dropout, which still face challenges such as inadequate explanation and low prediction accuracy. In view of these problems, this paper proposes a MOOC dropout prediction method using learning process model and LightGBM algorithm. This method utilizes the learning process model to analyze learning behavior and generate feature vectors to provide a clear interpretation. Based on these feature vectors, the LightGBM algorithm is adopted to predict dropout. Compared to the related methods, the dropout prediction method proposed in this paper demonstrates improvements of 1.58%, 1.2%, 1.1%, and 0.864% in Recall, F1, Precision, and AUC respectively.