A Deep learning Grade Prediction Model of Online Learning Performance Based on knowledge learning representation

In recent years, due to the impact of Coronavirus disease (COVID-19), digital platforms have developed rapidly and accumulated a large amount of data. To better utilize the comprehensive and diverse data stored in online platforms for data mining, such as learning behavior analysis or performance prediction, and to provide guidance and valuable feedback for educator became more important. For the current analysis of learning behaviors by time series data with DNN method, the interpretability is not enough. This paper proposes a method based on the simultaneous use of learning behaviors and learning materials to obtain the representation of learned knowledge, and through multiple cross-validations, the effect of this knowledge representation has a certain improvement on the original data, and the interpretability can promote the feedback function.