Research on Personalized Exercise Recommendation Based on Graph Representation Learning

Abstract

Traditional exercise recommendation systems often ignore the cognitive differences of learners and the dependencies between exercises, resulting in a mismatch between the recommended exercises and the learners' learning level and needs, which affects the recommendation effect. To solve this problem, a graph recognition system A personalized exercise recommendation algorithm for knowledge diagnosis is proposed in this paper. Firstly, build and use the learners' answer records to dig out the relationship between the knowledge concepts, and build a domain knowledge concept map. Then, based on graph cognitive diagnosis, the learner's mastery level of each knowledge concept (Abbreviated as KC) is calculated and learners' knowledge structure diagram is obtained. Then, graph representation learning is used to recommend personalized exercises for learners based on diagnosis results. Finally, the effectiveness of the personalized exercise recommendation model and heterogeneous graph representation learning proposed in this paper are verified through simulation experiments on public datasets. Compared with traditional methods, this method has achieved significant improvements in recommendation accuracy and learning effect.