Predicting students at risk of academic failure using learning analytics in the learning management system

Online learning platforms have become commonplace in modern society today, but high dropout rates and decrement students’ performance still require more attention in such online learning environments. The purpose of this research is to accelerate the identification of students at risk of academic failure in order to take appropriate corrective action. Therefore, we have proposed model to achieve this goal and ultimately improve the performance of students and faculty. Then, for early prediction of students at risk of academic failure, the short-term memory neural network (LSTM) and the widely used support vector algorithm have been used to analyze students’ time based behaviors using data from the University of Tehran e-learning system. To demonstrate the optimal performance of the predictive algorithm, we compared the LSTM network with the support vector algorithm with different evaluation criteria. The results show that the use of LSTM network for early prediction of students at risk provides higher predictive accuracy compared to the support vector machine algorithm. In this research, our method in predicting students’ performance with LSTM network has achieved 94% accuracy and with support vector machine algorithm has achieved 88% accuracy. In addition, the Area Under the Curve (AUC) was 0.936 and 0.882, respectively, using the LSTM algorithm and the support vector machine. Therefore, according to the obtained results, it can be seen that our proposed algorithm has an important and effective contribution to improving the final performance of teachers and students during the course.