Research on knowledge tracing based on learner fatigue state

Abstract

Knowledge tracing aims to predict how learners will perform in future exercises on related concepts and to track changes in their knowledge state. Existing models have not fully considered the physical and mental fatigue that occurs in learners during prolonged learning tasks, which leads to reduced problem-solving ability and affects their learning efficiency and performance. This article proposes Attention-Centric Knowledge Tracing to address the above issues. This method combines the Grit theory to evaluate the learner’s fatigue state and explores the potential impact of learning tasks on the learner’s fatigue state through deep graph convolutional networks. In particular, this article employs a multilayer perceptual network with scaled dot-product attention to process information dynamically, focusing on the critical information the learner needs at a given moment and effectively incorporating it into the knowledge framework. This article compared the fourteen knowledge tracing models in the experiment to the two benchmark data sets. The results indicate that knowledge tracing in the center of attention outperforms the baseline model in predicting learners’ future responses.