Dynamic Key-Value Gated Recurrent Network for Knowledge Tracing

Knowledge tracing model is one of the important research fields to realize personalized intelligent education. The focus of the model is to trace students' knowledge mastery from the records of students' answering performance. However, the existing knowledge tracing models don't consider the problems of students' knowledge application ability and forgetting rules. Based on the external memory mechanism of Dynamic Key-Value Memory Networks (DKVMN), this paper proposes a knowledge tracing model based on a dynamic key-value gated recurrent network (DKVGRU), which designs a knowledge update network inspired the idea of Gated Recurrent Unit (GRU). DKVGRU calculates the proportion of concepts students apply and measures the degree of forgetting of learned concepts, which traces the knowledge state of each concept well. In this paper, the area under the receiver operating characteristic curve (AUC) of the prediction result is used as an evaluation indicator. The performance of DKVGRU is higher than DKVMN on four public datasets.