Relief Approach for Predicting Learner Performance on MOOC
Abstract— The advent of Massive Online Open Courses(MOOC) has led to the availability of large educational datasets collected from researchers. The problem of predicting learner performance on MOOC has received much attention. Considering improving the predictive effect of the learner performance on MOOC, we propose Relief approach to select feature of learners. Based on the online data of edX platform, we divide the characters of learners into three categories, use Relief algorithm to select seven important features, and adopt several classical supervised machine learning methods to build the model for learner performance prediction. The experiments show that learner performance is mainly determined by two kinds of characters of learner type and learner behavior. The Logistic Regression algorithm and Support Vector Machine algorithm is demonstrated that they have high accuracy in predicting learner performance by comparison of the evaluation metrics.
Index Terms— Data Mining, MOOC, Prediction
College of Computer Engineering in Bengbu University of China, CHINA
Cite: Cheng Ma, "Relief Approach for Predicting Learner Performance on MOOC," Proceedings of 2018 the 8th International Workshop on Computer Science and Engineering, pp. 323-329, Bangkok, 28-30 June, 2018.