DKVMN-KAPS: Dynamic Key-Value Memory Networks Knowledge Tracing With Students' Knowledge-Absorption Ability and Problem-Solving Ability

Abstract

Knowledge tracing aims to predict students' future question-answering performance based on their historical question-answering records, but the current mainstream knowledge tracing model ignores the individual differences in different students' knowledge-absorption and problem-solving abilities, which leads to a poor prediction of students' question-answering performance by the model. To solve this, Dynamic Key-Value Memory Networks Knowledge Tracing with Students' Knowledge-Absorption Ability and Problem-Solving Ability (DKVMN-KAPS) is proposed in this paper. Firstly, a hierarchical convolutional neural network is used to consider students' knowledge mastery at multiple time steps, and then quantify students' knowledge-absorption ability, aiming to more accurately portray students' knowledge states; secondly, an autoencoder is used to dynamically update students' problem-solving ability at each time step; and finally, students' question answering performance is predicted by considering the students' knowledge state, problem-solving ability, and question features. Extensive experiments on three datasets show that the prediction performance of DKVMN-KAPS outperforms existing models and improves the prediction accuracy of deep knowledge tracing models.