ASRCD: Adaptive Serial Relation-Based Model for Cognitive Diagnosis

Zhuonan Liang, Dongnan Liu, Yuqing Yang, Caiyun Sun, Weidong Cai, Peng Fu

Abstract: Cognitive diagnosis (CD) is a critical task in the education field, aimed at assessing the true concept proficiency of learners. Recent studies have highlighted the significance of concept relations (e.g., concept Addition and concept Multiplication in mathematics) in CD. While advanced research has contributed to concept relation modeling, there remains a gap in automatic building and adaptive integration of relation modeling. To address these challenges, we present an innovative approach called the Adaptive Serial Relation-based model for Cognitive Diagnosis (ASRCD). Our method begins by constructing a Concept Serial Relation Graph (CSRG) to automatically mine concept relations from learner response sequences. Next, a refined graph attention network (GAT) is designed to weight the concept relations for effective aggregation. Finally, we establish a comprehensive CD model that incorporates concept relations, leveraging the extendibility of CSRG, allowing it to be seamlessly integrated into various existing CD methods. To evaluate the performance of our model, we conduct experiments on two real-world datasets from education practice. The experimental results demonstrate that our proposed ASRCD model achieves outstanding accuracy and exhibits excellent extendibility, showcasing its effectiveness and potential in enhancing cognitive diagnosis tasks.

External IDs:dblp:conf/iconip/LiangLYS0F23