Deep Knowledge Tracking based on Attention Mechanism for Student Performance Prediction

Due to the widespread application of online tutoring systems today, researchers have developed a strong interest in using online tutoring systems to improve students' learning performance. Researchers can use deep knowledge tracking methods to model students' learning performance based on their historical learning records, recommend learning resources to students based on their learning performance, and develop personalized learning strategies. At present, some researches on deep knowledge tracing using exercise text information and have achieved good results. But this method needs to rely on the complete text of the exercise, it will cause problems in difficult to extract the concept of the exercise, and due to the reason of intellectual property and privacy protection complete exercise texts are difficult to obtain. Therefore, in this study, we use a small amount of concept information, combined with the attention mechanism, to improve deep knowledge tracking, and to promote the ability of prediction student learning performance. The experimental results show that the method proposed in this paper has some improvements in AUC and ACC.