DRG: A dual relational graph framework for course recommendation.

Area of Science:

• Educational Technology
• Artificial Intelligence
• Data Science

Background:

• Course recommendation systems are crucial for personalized learning and improving teaching quality.
• Large Language Models (LLMs) show promise but struggle with data sparsity.
• Data sparsity limits the accuracy of traditional and LLM-based recommendation models.

Purpose of the Study:

• To propose a Dual Relationship Graph (DRG) framework to address data sparsity in course recommendation.
• To model both course-course and user-course relationships for enhanced recommendation accuracy.
• To develop a scalable and effective solution for personalized course recommendations in sparse educational environments.

Main Methods:

• Constructing a course-based graph using LLM semantic reasoning, collaborative filtering, clustering, and association rule mining.
• Building a user-based graph via collaborative filtering and LLM preference inference.
• Integrating dual graphs through joint learning and collaborative reasoning within a unified pipeline.

Main Results:

• DRG framework significantly alleviated data sparsity, increasing link coverage by 37.88% and 12.67% on two datasets.
• DRG demonstrated superior performance in task ranking compared to single-relationship approaches.
• The proposed DRG module enhanced both traditional and LLM-based recommendation systems.

Conclusions:

• The Dual Relationship Graph (DRG) framework effectively addresses data sparsity in educational recommendation systems.
• Modeling dual relationships and integrating LLM-driven semantic understanding leads to improved recommendation accuracy.
• DRG is a versatile, plug-and-play module that enhances existing recommendation models and offers a scalable solution.