Social Trust Prediction Using Heterogeneous Networks.

Area of Science:

• Social Network Analysis
• Data Mining
• Information Retrieval

Background:

• Online users increasingly rely on trust information for decision-making on social websites.
• Social network data often suffer from sparsity, limiting information availability and trust prediction accuracy.
• Traditional trust prediction methods focus solely on graph topology, neglecting valuable ancillary information.

Purpose of the Study:

• To address data sparsity and improve trust prediction in social networks.
• To propose a novel method for aggregating heterogeneous social networks for enhanced trust prediction.
• To leverage user-group-level similarity and shared patterns across multiple social networks.

Main Methods:

• Developed a joint social networks mining (JSNM) method for correlated graph analysis.
• Employed a joint learning model to simultaneously learn individual graph structures and explore inter-graph similarities.
• Utilized an alternative technique and auxiliary functions for optimizing the objective function with proven convergence.

Main Results:

• The JSNM method significantly improves trust prediction accuracy in the target graph.
• The approach enhances other information retrieval tasks in auxiliary graphs by utilizing shared structures.
• Extensive experiments on synthetic and real-world data validate the effectiveness of the proposed method.

Conclusions:

• Aggregating heterogeneous social networks with JSNM is effective for trust prediction.
• The method successfully transfers knowledge and patterns across correlated social networks.
• JSNM offers a robust solution for enhancing trust prediction and information retrieval in sparse social network data.