Privacy-preserving item-based recommendations over partitioned data with overlaps

 

Ibrahim Yakut  and Jaideep Vaidya, 2017.

International Journal of Business Information Systems Vol. 25, No. 3

 

User ratings are essential for recommender systems, and when ratings are insufficient, companies might use partitioned data to operate these systems. Privacy-preserving schemes enable this, but they often overlook potential rating overlaps in partitioned data. These overlaps can complicate collaboration, especially when the values differ. This study examines this divergent rating issue over arbitrarily partitioned data (APD) (Yakut and Polat, 2024) and proposes new schemes. These schemes allow two parties to perform item-based collaborative filtering on partitioned data with different overlaps, improving prediction quality with manageable overheads. 

       

In the context of overlapped ratings, PPCF over partitioned data has already investigated by Memis and Yakut (2014) and they propose PPCF solutions with the assumption of overlapped ratings held by both party are the same, i.e. they are duplicate ratings. They offer alternative solutions operating with ignoring and handling overlaps and they compare both alternatives. However, there are still challenges with overlaps and overlapped ratings may not be duplicates as in (Memis and Yakut, 2014) For instance, there are also open questions: how parties can get rid of such divergent values? What are their effects on prediction quality? In this study we examined these questions and propose solutions in this context. We set up emprical analysis and discussed simulation results. You can access our full article from Inderscience Enterprises Ltd. Web Site:

https://www.inderscienceonline.com/doi/epdf/10.1504/IJBIS.2017.084449

 

References

Yakut, I. and Polat, H. (2012) “Arbitrarily distributed data-based recommendations with privacy,” Data & Knowledge Engineering, vol. 72, pp. 239-256.

Memis, B. and Yakut, I. (2014) “Privacy-Preserving Two-Party Collaborative Filtering on Overlapped Ratings,” KSII Transactions on Internet and Information Systems, vol. 8, no. 8, pp.
2948-2966.