Exponent-Enhanced Attentive Knowledge Tracing based Online Learning Reinforcing
Hao Wu, Xu Bin, Yuekang Cai · 2021
As an important component of online learning education systems, knowledge tracing (KT) has been the focus of numerous researches at present. However, most of the existing KT methods evaluate the relevance between exercises without paying attention to the association among exercises involving multiple concepts. A novel Exponent-Enhanced Attentive Knowledge Tracing (EEAKT) method is proposed to solve multi-concept problems, which integrates the attention mechanism and the conceptual layer of exercises. Based on the related analysis, the concept information through the nonlinear change of the exponential function will be identified when the concept is complex. The identified exercises are reinforced by attention mechanism. Extensive experiments on three real-world datasets show that EEAKT outperforms existing KT methods. Moreover, EEAKT can automatically identify the knowledge concepts that students need to update, which consolidate individual memory.