A Knowledge Tracing Model Based on Hierarchical Heterogeneous Graphs
Bin Li, Yeh-Cheng Chen, Hongle Du, Y. Zhang
Mathematics
Problems Identified (4)
Knowledge-space relationship modeling: Knowledge tracing must account for complex relationships among learners, exercises, and knowledge concepts that affect learning outcomes.
Knowledge-concept interdependency modeling: Learner performance is influenced by mastery of relevant concepts and interdependencies among those concepts.
Knowledge-space relationship modeling: Knowledge tracing must account for complex relationships among learners, exercises, and knowledge concepts that affect learning outcomes.
Knowledge-concept interdependency modeling: Learner performance is influenced by mastery of relevant concepts and interdependencies among those concepts.
Proposed Solutions (4)
HHGKT hierarchical heterogeneous graph: The study proposes a Hierarchical Heterogeneous Graph Knowledge Tracing model that represents learners, exercises, and knowledge concepts in a hierarchical heterogeneous graph.
Concept-interdependency graph encoding: The model incorporates interdependencies among knowledge concepts into the graph structure while capturing learner–concept and exercise–concept interactions.
HHGKT hierarchical heterogeneous graph: The study proposes a Hierarchical Heterogeneous Graph Knowledge Tracing model that represents learners, exercises, and knowledge concepts in a hierarchical heterogeneous graph.
Concept-interdependency graph encoding: The model incorporates interdependencies among knowledge concepts into the graph structure while capturing learner–concept and exercise–concept interactions.
Results (3)
Knowledge-space complexity representation:
Concept-interdependency accuracy gain:
Heterogeneous graph accuracy gain:
Research Domain
Intelligent tutoring systems; knowledge tracing; educational data mining