Calibrated Recommendation Algorithm for Multi-Source Heterogeneous Online Learning Resources: A Dual Knowledge Graph Approach

Online learning recommendation is critical for talent cultivation and high-quality education. However, existing algorithms often prioritize personalization over learning goals, leading to issues such as "cold start" and "data sparsity" that hinder effective resource recommendation. This study proposes the SKM algorithm, a goal-oriented learning recommendation algorithm for multi-source heterogeneous e-learning resources. SKM leverages a dual knowledge graph to align recommendations with learning objectives and integrates a graph label-connected cluster clustering algorithm (KGLC) to enhance recommendation effectiveness. Evaluated across four types of datasets, SKM demonstrates superior performance, achieving a 3.9 times higher correction rate compared to mainstream algorithms. The KGLC algorithm further improves SKM’s correction effect by efficiently clustering similar labels. The proposed SKM and KGLC algorithms address key limitations in existing recommendation systems, providing a robust solution for online learning resource recommendation. This work contributes to advancing educational technology and lays a foundation for future research in goal-oriented learning systems.