Knowledge Tracing with Multi-Feature Fusion and Question Difficulty
Haikun Liu, Weijie Peng, Chao Qi · 2022
Knowledge Tracing (KT) aims to trace students’ knowledge state and predict their performance on questions in online tutoring system. For the moment, most methods assume that all questions are equal to every student, that is to say, question itself has not any effect on students’ knowledge state. And they have not considered the effect of question difficulty (QD) on students’ knowledge state. To address this issue, some researchers have worked on the effect of question difficulty on students’ knowledge state. But, they have not considered the effect of question difficulty and knowledge concept difficulty (KCD) on students’ knowledge acquisition ability and knowledge state. In this paper, we propose a novel model called Question Difficulty Estimation Knowledge Tracing (QDEKT) model, which comprehensively considers question difficulty and knowledge concept difficulty and introduces students’ real response and predicted response on questions into the model, to evaluate students’ knowledge acquisition ability and dynamically update students’ knowledge state. Finally, extensive experiments show the effectiveness of our model.