Cooperation Learning From Multiple Social Networks: Consistent and Complementary Perspectives.

Area of Science:

• Social Network Analysis
• Data Fusion
• Machine Learning

Background:

• Internet users actively engage with an average of 2.82 social media accounts.
• Users express diverse views across multiple social platforms, creating fragmented user data.
• Existing methods often overlook the cooperative relationships between different data sources.

Purpose of the Study:

• To develop a unified model for fusing social signals from multiple social networks.
• To address the challenges of source consistency, complementarity, and confidence in multi-network data fusion.
• To enhance downstream analyses, such as user interest inference, by effectively integrating user data.

Main Methods:

• Proposed a novel unified model incorporating co-regularization for source consistency, complementarity, and confidence.
• Derived the theoretical solution for the proposed model.
• Validated the model through a real-world user interest inference application.

Main Results:

• The unified model effectively fuses social signals by considering source consistency, complementarity, and confidence.
• Experimental results demonstrate the superiority of the proposed model over state-of-the-art competitors.
• The model significantly improves the performance of user interest inference tasks.

Conclusions:

• The proposed unified model offers a robust approach for multi-network social signal fusion.
• Co-regularizing source consistency, complementarity, and confidence is crucial for effective data integration.
• This work advances the field of user characterization by leveraging diverse social media data.