Deep Knowledge Tracking Method Based on DKVTMN-DTCN Model

Abstract

With the rapid development of deep neural networks, deep knowledge tracking models have become one of the most important research areas in educational data mining. While most knowledge tracking models assume that the forgetting level of all students is constant, in reality, forgetting levels are affected by time intervals and different learning abilities. The learning ability of the same student changes over time, and there is variability in the learning ability of different students, so students will forget their knowledge to different degrees within the same time interval of answering questions. The DKVMN-DT uses CART to preprocess student behavioral characteristics. However, the time series analysis capability of the decision tree model is very limited, and the prediction results it obtains are basically within a certain training range, so for features with the significant trend, the decision tree cannot directly predict the time series changes in student behavior. In summary, a DKVTMN-DTCN knowledge tracking model that combines a temporal forgetting mechanism based on a priori student ability and automatically extracts temporal features using temporal convolutional networks is proposed to capture students’ long-term behavioral characteristics and personalize knowledge forgetting. Experimental results show that the prediction performance of the DKVTMN-DTCN is significantly improved compared with the classical model on both datasets.