p02 / Journal
FedPOIRec: Privacy-Preserving Federated POI Recommendation with Social Influence
Information Sciences, 623, 767-790 · 2023 · Online 2022
In plain language
FedPOIRec recommends places through federated learning and incorporates social influence without pooling users’ location histories.
Research context
Personalization can use social knowledge while keeping raw location records on their owners’ devices.
Approach
Local recommendation models, blind update aggregation, and CKKS homomorphic encryption for integrating friends’ learned preferences.
Findings
Across five real datasets and two recommendation models, the preprint reports quality comparable to centralized approaches with low client-side overhead for social integration.
Summary based on abstract. The journal issue is dated April 2023; the article was first available online in December 2022. Summary follows the author preprint abstract.