Exercise recommendation method based on knowledge tracing and concept prerequisite relations

With the development of technology, the teaching environment has changed greatly. As an educational resource, exercise plays an important role in students' personalized learning service. Therefore, how to recommend appropriate exercises to students is a key problem to be solved urgently. The exercise recommendation method analyses students' history answer sequences and provides personalized exercise recommendation service for students. Previous exercise recommendation methods assume that students' knowledge states are fixed, so these methods cannot recommend exercise according to the changes of student ability. In addition, the existing methods do not take concept prerequisite relations into consideration. In this paper, we propose Exercise Recommendation method based on Knowledge Tracing and Concept Prerequisite relations (ER-KTCP). Firstly, ER-KTCP can capture the changes of students' knowledge states. Secondly, ER-KTCP can adjust the details of recommendation strategy according to the changes of students' knowledge states. Thirdly, ER-KTCP recommends exercises according to the relations between concepts and the difficulty of exercises, which makes selected exercises more reasonable. Besides, we propose a new metric to evaluate the improvement of student's score after he has done the recommended exercises. Experiments on multiple data sets show that ER-KTCP performs better in exercise recommendation than state-of-the-art methods.