Deep Attentive Model for Knowledge Tracing

Knowledge Tracing (KT) is a crucial task in the field of online education, since it aims to predict students' performance on exercises based their learning history. One typical solution for knowledge tracing combine classic models educational psychology, such as Item Response Theory (IRT) and Cognitive Diagnosis (CD), with Deep Neural Networks (DNN) technologies. In this solution, student related are mapped into feature vectors student's at current time step, however, does not consider impact historical behavior sequences, relationships between sequences students. paper, we develop DAKTN, novel model which assimilates tackle challenge better tracing. To be specific, apply pooling layer incorporate sequence embedding layer. After that, further design local activation unit, can adaptively calculate representation by taking relevance consideration respect candidate exercises. Through experimental results three real-world datasets, DAKTN significantly outperforms state-of-the-art baseline models. We also present reasonableness ablation testing.