Trustworthy End-to-End Deep Student Knowledge Portrait Modelling Method

Student knowledge portrait means a comprehensive and accurate representation of student’s mastery for knowledge concepts. Generally, knowledge tracing methods are used in intelligent education to summarize and predict potential students’ mastery of knowledge concepts based on students’ learning data. However, the prediction of the knowledge tracing method is inconsistent with the student knowledge portrait, leading to not credible portrait. In this paper, we first define the student knowledge portrait task based on the end-to-end object of students’ knowledge mastery, and then propose a novel deep knowledge portrait (DKP) model. Specifically, we first represent the learning interaction with difficulty and knowledge concepts to distinguish learning interactions on the knowledge level. Besides, we adopt the bidirectional long and short time memory network (Bi-LSTM) to trace the student’s knowledge states based on the learning record. Finally, we predicted the student knowledge portrait using a multi-head attention pooling layer to focus on the historical knowledge states related to the knowledge concepts to predict the mastery. Experimental results on three real datasets show that the proposed method is more suitable for student knowledge portrait task so as to obtain a more trustworthy student knowledge portrait, and overperforms the baselines on metrics.