Deep Knowledge Tracing Integrating Multiple Learning Behaviors

Abstract

By analyzing students' external learning behaviors, knowledge tracing quantifies students' latent knowledge state on this learning task, so as to further develop targeted learning and teaching plans and promote personalized learning. Students' learning behaviors in online learning platforms are diverse, such as exercise, exam and tutorial browsing. However, most of the existing knowledge tracing models only consider exercise and do not fully utilize other behaviors that also reflect students' learning process. In order to solve this problem, this paper proposes a deep knowledge tracing with multiple learning behaviors model (DKT-MLB), which combines multiple learning behaviors with knowledge concepts. The effectiveness of the proposed model is verified by experiments in a dataset built in real online learning platforms.