Predicting Student Achievement Based on Temporal Learning Behavior in MOOCs

Abstract

With the development of data mining technology, educational data mining (EDM) has gained increasing amounts of attention. Research on massive open online courses (MOOCs) is an important area of EDM. Previous studies found that assignment-related behaviors in MOOCs (such as the completed number of assignments) can affect student achievement. However, these methods cannot fully reflect students’ learning processes and affect the accuracy of prediction. In the present paper, we consider the temporal learning behaviors of students to propose a student achievement prediction method for MOOCs. First, a multi-layer long short-term memory (LSTM) neural network is employed to reflect students’ learning processes. Second, a discriminative sequential pattern (DSP) mining-based pattern adapter is proposed to obtain the behavior patterns of students and enhance the significance of critical information. Third, a framework is constructed with an attention mechanism that includes data pre-processing, pattern adaptation, and the LSTM neural network to predict student achievement. In the experiments, we collect data from a C programming course from the year 2012 and extract assignment-related features. The experimental results reveal that this method achieves an accuracy rate of 91% and a recall of 94%.