Hybrid Recommendation Network Model with a Synthesis of Social Matrix Factorization and Link Probability Functions.

Area of Science:

• Artificial Intelligence
• Machine Learning
• Data Science

Background:

• Recommender systems are crucial for daily decisions but suffer from sparsity issues, limiting recommendation quality.
• Existing models struggle with sparse data, impacting user experience in areas like music discovery.

Purpose of the Study:

• To introduce a novel hybrid recommendation model, Relational Collaborative Topic Regression with Social Matrix Factorization (RCTR-SMF), for enhanced music artist recommendations.
• To leverage auxiliary domain knowledge, social network information, and item relational structures to overcome sparsity and cold-start problems.

Main Methods:

• Developed a hierarchical Bayesian hybrid model, RCTR-SMF, integrating Social Matrix Factorization and Link Probability Functions with Collaborative Topic Regression.
• Utilized item content, user-item interactions, social networking data, and item-relational network structures for prediction.

Main Results:

• The RCTR-SMF model demonstrated superior performance on a large real-world social media dataset.
• Achieved a recall of 57%, outperforming other state-of-the-art recommendation algorithms.

Conclusions:

• RCTR-SMF effectively addresses sparsity and cold-start challenges in recommender systems.
• The model's integration of diverse data sources leads to more accurate and reliable recommendations, particularly in music artist suggestions.