Deep learning knowledge tracing based on behavioral features
Shuyan Wang, Yang Wang · 2023
Knowledge tracing assesses students’ understanding of knowledge concepts and learning progress by tracking their learning behaviors and performance in answering questions. Research has shown that existing deep learning-based knowledge tracing models only consider the training and outcomes of students as inputs, ignoring the rich behavioral information generated by students during the learning process. This information can provide a clearer understanding of students’ mastery of the knowledge points. In this study, a deep learning knowledge tracing model based on a dynamic key-value memory network model with strong interpretability is selected as the foundation. The XGBoost decision tree algorithm is used to preprocess the student’s behavioral information, capturing the impact of student behavioral factors on their cognitive states, and integrating the classification results with the learning features. Experimental results demonstrate that the improved deep learning knowledge tracing model outperforms traditional deep learning knowledge tracing models in terms of predictive accuracy on educational datasets.