基于图神经网络的深度知识追踪方法综述OA

With the advancement of digital transformation in education,deep knowledge tracing has emerged as a research focus in intelligent education due to its ability to dynamically model learners'knowledge states and support personalized learning through advanced algorithms.However,traditional deep learning-based models often struggle to capture the com-plex relationships between learners and knowledge concepts.Deep knowledge tracing based on graph neural networks address this limitation by leveraging graph structures to model these intricate interactions more effectively,leading to improved prediction accuracy and interpretability.This paper systematically reviews recent research on deep knowledge tracing based on graph neural networks.It introduces the concepts of knowledge tracing and graph neural networks.Then,it summarizes the deep knowledge tracking methods based on graph neural networks,graph convolutional networks,graph attention networks,and graph memory networks,and categorizes existing methods into three perspectives:using dif-ferent graph structures,incorporating additional features,and integrating educational psychology factors.In addition,the paper reviews five benchmark datasets commonly used in deep knowledge tracing and presents performance evaluations of representative models on these datasets.Finally,it analyzes the existing challenges and open issues in this research domain.