Leveraging Leiden communities for enhanced collaborative filtering with matrix factorization techniques.

Area of Science:

• Computer Science
• Data Science
• Artificial Intelligence

Background:

• Personalized recommendation systems are crucial for user experience.
• Collaborative filtering and matrix factorization are common techniques.
• Scalability and accuracy challenges exist with large, complex datasets.

Purpose of the Study:

• To propose a novel strategy combining matrix factorization with Leiden community detection.
• To enhance the scalability and quality of personalized recommendations.
• To address limitations of existing recommendation system approaches.

Main Methods:

• Representing rating data as a bipartite graph.
• Applying Leiden community detection to identify user/item groups.
• Utilizing matrix factorization (MF, SVD++, FANMF) within detected communities.
• Consolidating predictions and evaluating performance metrics (RMSE, MSE, MAE).

Main Results:

• Significant improvements in recommendation effectiveness.
• Noticeably reduced error values (RMSE, MSE, MAE).
• Enhanced computational efficiency compared to traditional methods.

Conclusions:

• The proposed hybrid approach effectively scales recommendation systems.
• Combining community detection with matrix factorization offers a robust solution.
• This methodology improves both accuracy and efficiency in personalized recommendations.