Privacy-preserving recommender system using the data collaboration analysis for distributed datasets.

Area of Science:

• Computer Science
• Data Privacy
• Machine Learning

Background:

• High-quality recommendations require integrating distributed datasets from multiple parties.
• Sharing distributed datasets raises concerns about protecting personal and confidential information.
• Existing recommender systems often lack robust privacy-preserving mechanisms for data collaboration.

Purpose of the Study:

• To establish a framework for privacy-preserving recommender systems using collaborative analysis of distributed datasets.
• To demonstrate the effectiveness of privacy-preserving techniques in improving recommendation accuracy.
• To enable secure data sharing and integration for enhanced user recommendations.

Main Methods:

• Development of a novel framework for privacy-preserving recommender systems.
• Utilizing data collaboration analysis on distributed datasets.
• Conducting numerical experiments with two public rating datasets.

Main Results:

• The proposed privacy-preserving method improved rating prediction accuracy for distributed datasets.
• Prediction errors were reduced by an average of 4.5% and up to 7.0% compared to individual dataset analysis.
• The framework successfully protected personal and confidential information during data integration.

Conclusions:

• Collaborative analysis of distributed datasets within a privacy-preserving framework enhances recommender system performance.
• The developed method offers a viable solution for secure data sharing in recommendation tasks.
• This research opens new avenues for privacy-preserving techniques in the field of recommender systems.