Author ORCID Identifier
Date Available
6-16-2023
Year of Publication
2023
Degree Name
Doctor of Philosophy (PhD)
Document Type
Doctoral Dissertation
College
Engineering
Department/School/Program
Computer Science
First Advisor
Dr. Zongming Fei
Abstract
Location-based services (LBS) have become increasingly popular, with millions of people using mobile devices to access information about nearby points of interest (POIs). Personalized POI recommender systems have been developed to assist users in discovering and navigating these POIs. However, these systems typically require large amounts of user data, including location history and preferences, to provide personalized recommendations.
The collection and use of such data can pose significant privacy concerns. This dissertation proposes a privacy-preserving approach to POI recommendations that address these privacy concerns. The proposed approach uses clustering, tabular generative adversarial networks, and differential privacy to generate synthetic user data, allowing for personalized recommendations without revealing individual user data. Specifically, the approach clusters users based on their fuzzy locations, generates synthetic user data using a tabular generative adversarial network and perturbs user data with differential privacy before it is used for recommendation.
The proposed approaches achieve well-balanced trade-offs between accuracy and privacy preservation and can be applied to different recommender systems. The approach is evaluated through extensive experiments on real-world POI datasets, demonstrating that it is effective in providing personalized recommendations while preserving user privacy. The results show that the proposed approach achieves comparable accuracy to traditional POI recommender systems that do not consider privacy while providing significant privacy guarantees for users.
The research's contribution is twofold: it compares different methods for synthesizing user data specifically for POI recommender systems and offers a general privacy-preserving framework for different recommender systems. The proposed approach provides a novel solution to the privacy concerns of POI recommender systems, contributes to the development of more trustworthy and user-friendly LBS applications, and can enhance the trust of users in these systems.
Digital Object Identifier (DOI)
https://doi.org/10.13023/etd.2023.270
Recommended Citation
Cui, Longyin, "PERSONALIZED POINT OF INTEREST RECOMMENDATIONS WITH PRIVACY-PRESERVING TECHNIQUES" (2023). Theses and Dissertations--Computer Science. 136.
https://uknowledge.uky.edu/cs_etds/136
