Recommendation based on trust diffusion model.

Area of Science:

• Computer Science
• Artificial Intelligence
• Information Retrieval

Background:

• Recommender systems are crucial for personalized online information delivery.
• Traditional systems face challenges like the cold start and data sparsity problems, limiting their efficiency.
• Existing solutions often lack satisfactory performance and efficiency.

Purpose of the Study:

• To propose an efficient recommender system model that overcomes the limitations of traditional approaches.
• To leverage implicit trust relationships within a network to alleviate data sparsity.
• To enhance recommendation accuracy by integrating user preferences with trusted connections' interests.

Main Methods:

• The proposed method integrates the trust diffusion (DiffTrust) algorithm with probabilistic matrix factorization (PMF).
• DiffTrust analyzes trust diffusion across users, utilizing implicit network relationships to mitigate data sparsity.
• PMF combines individual user tastes with the interests of their trusted connections.

Main Results:

• The DiffTrust + PMF model was evaluated on the Flixster, Moviedata, and Epinions datasets.
• Experimental results demonstrate superior performance in recommendation accuracy.
• Key performance metrics, including Root Mean Square Error (RMSE), Recall, and F Measure, showed significant improvements.

Conclusions:

• The combined DiffTrust + PMF model offers an effective solution for enhancing recommender system performance.
• This approach successfully addresses the data sparsity and cold start problems in recommendation.
• The model provides a robust and efficient method for personalized information delivery through trust networks.