Enhancing Recommender Systems through Imputation and Social-Aware Graph Convolutional Neural Network.

Area of Science:

• Computer Science
• Artificial Intelligence
• Data Science

Background:

• Recommendation systems are crucial for content discovery.
• Collaborative filtering methods analyze user-item interactions, often facing challenges with sparse data.
• Data sparsity hinders accurate and personalized recommendations.

Purpose of the Study:

• To develop a novel method for mitigating sparsity in recommendation systems.
• To improve the accuracy and personalization of recommendations by incorporating diverse data sources.

Main Methods:

• Incorporation of diverse data sources: trust statements and an imputation graph.
• Construction of an imputation graph based on user-item matrix and similar users' average rates.
• Utilization of a trust graph to capture user relationships and trust levels.
• Application of an attention mechanism to fine-tune the influence of combined graphs (user-item rating, trust, imputation).

Main Results:

• The proposed method demonstrates superior performance compared to state-of-the-art recommenders.
• Consistent outperformance observed in real-world dataset evaluations.
• Effective mitigation of sparsity challenges in recommendation systems.

Conclusions:

• The integrated approach effectively addresses sparsity in recommendation systems.
• The method offers enhanced personalization and effectiveness in content suggestions.
• This research contributes a robust solution for strengthening recommendation system performance.