Adaptive Exercise Recommendation Based on Cognitive Level and Collaborative Filtering

Abstract

Adaptive learning is inseparable from adaptive testing, and adaptive testing needs to rely on students cognitive level for personalized recommendations, and recommending exercises that match students’ cognitive level can enhance students’ learning interest and efficiency. Some of the existing exercise recommendation methods only recommend based on the similarity between students or exercises, while others only focus on students knowledge level for recommendation. The former ignores students’ cognitive level and tends to recommend exercises that are too difficult or too easy. The latter ignores the group nature among students, and the recommendation results are too homogeneous. To tackle these problems, in this work, we propose an adaptive exercise recommendation based on cognitive level and collaborative filtering (ACLCE). Learning students’ cognitive level by a deep knowledge tracking model, and then recommends exercises by a neural collaborative filtering algorithm combining students groupness. Extensive experiments on two-real world datasets show the effectiveness of ACLCE by significantly boosting the recommendation. Compared with the advanced baseline method, ACLCE achieves a significant improvement in recommendation effectiveness with the Six to eight percent performance gain.