Educational Data Mining: Dropout Prediction in XuetangX MOOCs

With the rapid development of educational data mining and learning analytics, this study tries to make sense of education data and improve teachers' competence and teaching experience. In recent years, massive open online courses (MOOCs) have become the first choice of online learning for tens of millions of people around the world. However, the dropout rates for MOOCs are high. The goal of dropout prediction is to predict whether learners will exhibit learning behavior in several consecutive days in the future. Therefore, in this study, we consider the correlation information of learners' learning behaviors for several consecutive days. Through the in-depth statistical analysis of learners' learning behavior, it is found that learners' learning behavior on the next day is similar to that of the previous day. Based on this characteristic, we propose a Lie group region covariance matrix to represent the local correlation information of learning behavior and construct a convolutional neural network model with a multidilation pooling module to extract the local correlation high-level features of learning behavior for dropout prediction. In addition, extensive experiments show that the local correlation of learners' learning behavior cannot be ignored, which is fully considered in our model. Compared with the existing methods, our method achieves the best experimental results in accuracy, F-measure, precision, and recall, which is better than the current methods.