Multi-Granularity Ensemble Interaction Graph Modeling for Knowledge Tracing

Jing Wang

doi:10.1016/j.knosys.2024.112834

Multi-Granularity Ensemble Interaction Graph Modeling for Knowledge Tracing

2025-01-01

0

PRE

AI

DOI：10.1016/j.knosys.2024.112834

Original

Original request for help

Save

Abstract

En 中文

Knowledge tracing (KT) is a crucial educational tool that models students' mastery of various knowledge concepts by analyzing their historical learning records. Mainstream studies have turned to Deep Neural Networks (DNNs) to effectively trace students' knowledge states. Nonetheless, these methods may encounter performance barriers due to two critical constraints: the treatment of learning records as singular time series data and the absence of accounting for interaction correlations. To overcome the limitations, we introduce a novel KT model, named M ulti-Granularity Ensemble I nteraction G raph Modeling for K nowledge T racing (MGIGKT). Our model incorporates two modules designed to model students' knowledge states from the perspectives of temporal perception and interaction dependency, respectively. This approach allows for a more comprehensive understanding of students' learning dynamics and the correlations between their interactions. We have conducted extensive experiments on four datasets to evaluate the superiority and effectiveness of our proposed MGIGKT model. The results of these experiments provide empirical evidence of the model's enhanced ability to trace students' knowledge states accurately.

Keywords:

Intelligent education

Knowledge tracing

Correctness prediction

Graph Neural Networks

Multi-Granularity Ensemble Interaction Graph Modeling for Knowledge Tracing

Abstract

Journal

Organization