Bidirectional Trust-Enhanced Collaborative Filtering for Point-of-Interest Recommendation.

Area of Science:

• Computer Science
• Artificial Intelligence
• Data Mining

Background:

• Personalized point-of-interest (POI) recommendation systems enhance user experience but face challenges like data sparsity and trustworthiness.
• Existing models often overlook the influence of location trust and context factors, and inadequately fuse user preferences with contextual information.

Purpose of the Study:

• To propose a novel bidirectional trust-enhanced collaborative filtering model for POI recommendation.
• To address trustworthiness by considering both user and location trust.
• To mitigate data sparsity using temporal, geographical, and textual content factors.

Main Methods:

• Developed a bidirectional trust-enhanced collaborative filtering model incorporating user and location trust.
• Introduced temporal factors for user trust and geographical/textual content factors for location trust.
• Employed weighted matrix factorization with POI category factors and a fused framework for integrating trust and preference models.

Main Results:

• The proposed model demonstrated significant improvements in recommendation performance.
• Achieved a 13.87% increase in precision@5 and a 10.36% increase in recall@5 on Gowalla and Foursquare datasets.
• Outperformed state-of-the-art recommendation models in extensive experiments.

Conclusions:

• The bidirectional trust-enhanced model effectively addresses trustworthiness and data sparsity in POI recommendation.
• The integration of user trust, location trust, and fused preference/context models leads to superior recommendation quality.
• The proposed approach offers a robust solution for enhancing personalized POI recommendation systems.