An efficient privacy-preserving point-of-interest recommendation model based on local differential privacy

Abstract With the rapid development of point-of-interest (POI) recommendation services, how to utilize multiple types users’ information safely and effectively for a better is challenging. To solve problems imperfect privacy-preserving mechanism insufficient response-ability complex contexts, this paper proposes hybrid POI model based on local differential privacy (LDP). Firstly, we introduce randomized response techniques k -RR RAPPOR disturb ratings social relationships, respectively propose virtual check-in time generation method deal with issue missing after disturbance. Secondly, simultaneously combining information, construct containing three sub-models. Sub-model 1 considers effect user preference, relationship, forgetting feature, trajectory similarity calculation. 2 analyzes geographical correlation POIs. 3 focuses categories Finally, generate results. test performance recommendation, design groups experiments real-world datasets comprehensive verifying. The experimental results show that proposed outperforms existing methods. Theoretically, our study contributes effective safe usage multidimensional data science analytics recommender system design. Practically, findings can be used improve quality services.