Attention-Based Hypergraph Knowledge Tracing

Knowledge tracing is a crucial aspect of intelligent education. It obtains the learner’s knowledge state by analyzing the historical interaction records of learners’ online responses. In past research, obtaining students’ knowledge state is a huge challenge in personalized education. In this paper, we propose an attention-based Hypergraph knowledge tracing method (HAGKT). The model is used to learn feature-knowledge state Hypergraphs by introducing Hypergraph representation learning. Specifically, by introducing the attention mechanism to learn the Hypergraph, the model can clearly discover the high-order relationships of the students’ knowledge state at each moment. We interpret the knowledge state at each moment as a weight matrix in the attention mechanism. Experiments on multiple real-world extensive datasets show that our proposed HAGKT outperforms state-of-the-art models.