DyGKT: Dynamic Graph Learning for Knowledge Tracing

Ke Cheng,Linzhi Peng,Pengyang Wang,Junchen Ye,Leilei Sun,Bowen Du

2024-07-30

Abstract:Knowledge Tracing aims to assess student learning states by predicting their performance in answering questions. Different from the existing research which utilizes fixed-length learning sequence to obtain the student states and regards KT as a static problem, this work is motivated by three dynamical characteristics: 1) The scales of students answering records are constantly growing; 2) The semantics of time intervals between the records vary; 3) The relationships between students, questions and concepts are evolving. The three dynamical characteristics above contain the great potential to revolutionize the existing knowledge tracing methods. Along this line, we propose a Dynamic Graph-based Knowledge Tracing model, namely DyGKT. In particular, a continuous-time dynamic question-answering graph for knowledge tracing is constructed to deal with the infinitely growing answering behaviors, and it is worth mentioning that it is the first time dynamic graph learning technology is used in this field. Then, a dual time encoder is proposed to capture long-term and short-term semantics among the different time intervals. Finally, a multiset indicator is utilized to model the evolving relationships between students, questions, and concepts via the graph structural feature. Numerous experiments are conducted on five real-world datasets, and the results demonstrate the superiority of our model. All the used resources are publicly available at <a class="link-external link-https" href="https://github.com/PengLinzhi/DyGKT" rel="external noopener nofollow">this https URL</a>.

Machine Learning

What problem does this paper attempt to address?

The paper aims to address several key issues in the field of Knowledge Tracing (KT), particularly the limitations of existing methods in handling student answer records. Specifically, the paper focuses on the following three dynamic characteristics and proposes corresponding solutions:

1. **Increasing Data Volume**: The number of student answer records continues to grow, while existing KT models only evaluate student states based on fixed-length learning sequences, which is inconsistent with real-world applications. The proposed method can adapt to infinitely growing data volumes.

2. **Changing Significance of Time Intervals**: Different time intervals have different impacts on students' learning states. Short time intervals may reflect students' proficiency, while long time intervals may indicate forgetting behavior. Existing KT models struggle to capture these different time dynamics. The paper distinguishes between short-term and long-term semantics through a dual-time encoding strategy, better describing students' behavior patterns.

3. **Complex Relationship Evolution**: The relationships between students, questions, and concepts are constantly changing. For example, students may repeatedly attempt the same question, and this repetitive behavior needs to be correctly identified. Existing methods find it difficult to handle such complex relationship structures. The paper introduces multi-set indicators to handle this complex evolving relationship.

To address the above issues, the authors propose a new model called DyGKT (Dynamic Graph-based Knowledge Tracing). This model uses Continuous-Time Dynamic Graphs (CTDG) to construct a dynamic knowledge tracing framework. DyGKT captures the interactions between students and questions by constructing continuous-time dynamic graphs and designs components such as dual-time encoding strategies and multi-set indicators to handle the dynamic changes in time and structure, respectively.

In summary, the goal of DyGKT is to improve existing knowledge tracing methods by considering the dynamic characteristics of student answer records, in order to more accurately predict students' future performance and recommend suitable learning materials.