On Deep Learning for Trust-Aware Recommendations in Social Networks.

Area of Science:

• Computer Science
• Artificial Intelligence
• Data Science

Background:

• Social network-based recommendation systems effectively address the cold-start problem for new users.
• User trust information is crucial for enhancing the reliability of recommendations.
• Matrix factorization (MF) is a dominant technique, but its performance depends heavily on initial feature vectors.

Purpose of the Study:

• To develop a novel trust-based recommendation approach for social networks.
• To leverage deep learning for initializing matrix factorization in trust-aware recommendations.
• To account for community effects within users' trusted friendships.

Main Methods:

• A two-phase recommendation process is proposed.
• Deep learning is utilized for initializing matrix factorization.
• User interests and trusted friends' interests are synthesized, incorporating community effects.

Main Results:

• The proposed approach demonstrates superior accuracy and effectiveness.
• Extensive experiments were conducted on real-world social network data.
• Performance was compared against state-of-the-art methods.

Conclusions:

• The novel trust-based approach enhances recommendation accuracy in social networks.
• Integrating deep learning for initialization and considering community effects improves performance.
• The method offers a significant advancement in trust-aware social recommendations.