Transfer collaborative filtering from multiple sources via consensus regularization.

Area of Science:

• Artificial Intelligence
• Computer Science
• Data Science

Background:

• Collaborative filtering is a key method for recommendation systems.
• Transfer learning enhances recommendations by incorporating external data.
• Existing methods often rely on single-source transfer, limiting potential.

Purpose of the Study:

• To develop a transfer collaborative filtering framework that leverages multiple data sources.
• To address the challenge of inconsistent information from diverse sources.
• To propose a novel consensus regularization technique for improved knowledge transfer.

Main Methods:

• Introduced the TRACER (TRA nsfer collaborative filtering via C onsE nsus R egularization) framework.
• Implemented consensus regularization to harmonize information from multiple sources.
• Developed an algorithm that learns and transfers knowledge concurrently.

Main Results:

• The TRACER framework effectively handles information inconsistency.
• Experimental results on real-world datasets demonstrate the algorithm's effectiveness.
• Simultaneous learning and transfer proved more efficient than sequential methods.

Conclusions:

• TRACER offers a robust solution for multi-source transfer learning in recommendation systems.
• Consensus regularization is a viable method for managing data heterogeneity.
• The proposed approach enhances the accuracy and utility of collaborative filtering.