Knowledge relation rank enhanced heterogeneous learning interaction modeling for neural graph forgetting knowledge...

Area of Science:

• Educational Data Mining
• Artificial Intelligence in Education
• Machine Learning for Learning Analytics

Background:

• Conventional knowledge tracing models often oversimplify exercise-knowledge relationships, leading to subjective bias and limited accuracy.
• Existing models struggle to capture the nuanced interactions between students, exercises, and skills.
• The Self-Attention Knowledge Tracing model highlights the importance of exercise-knowledge relationships but still has limitations.

Purpose of the Study:

• To propose a novel knowledge tracing model that addresses the limitations of existing methods.
• To mitigate subjective labeling bias in knowledge tracing.
• To enhance the accuracy of predicting student performance by modeling complex interactions.

Main Methods:

• Developed a Knowledge Relation Rank Enhanced Heterogeneous Learning Interaction Modeling for Neural Graph Forgetting Knowledge Tracing (KRR-HLIM-NGFT) model.
• Employed Graph Convolutional Networks (GCNs) to model intricate student-exercise-skill interactions.
• Utilized the Knowledge Relation Importance Rank Calibration method to fine-tune skill relation and Q-matrices, reducing bias.

Main Results:

• The proposed KRR-HLIM-NGFT model demonstrated superior performance compared to baseline models on two public datasets.
• The model showed enhanced accuracy across three key evaluation metrics.
• Fine-tuning matrices and GCNs effectively captured complex learning dynamics.

Conclusions:

• The novel knowledge tracing model significantly improves prediction accuracy in educational settings.
• The approach effectively reduces subjective bias inherent in traditional knowledge tracing methods.
• This work offers a more robust framework for understanding and modeling student learning processes.