Research On Point Of Interest Recommendation Preserving-privacy For Geosocial Networks

With the development of mobile Internet,point-of-interest recommendation provide great convenience to people's daily life in geosocial networks,which allows users to choose quickly and accurately from the huge volume of data.However,point-of-interest recommendation needs to mine the users' behavior information,which leads to user privacy leakage.Therefore,while enjoying points of interest recommendation service,how to protect user location privacy has become one of the most important issues.This thesis mainly focuses on the issue of users' location privacy disclosure in geosocial networks,which aims to improve the recommendation performance and protect users' location privacy.An improved recommendation algorithm is proposed to solve the low precision on the point-of-interest recommendation.Further,a privacy algorithm is proposed to resist location privacy exposure problem on point-of-interest recommendation.The specific research work is as follows:(1)In order to solve the problem of low recommendation accuracy in geosocial networks,a novel point-of-interest recommendation algorithm is proposed incorporating three factors:user preferences,social influence and distance weight.User preference is calculated by user history check-in records.The social influence is estimated by using the users' social link.Considering users' the current position,a term frequency inverse document frequency is used to predict the influence of distance weight on check-in.The experimental results show that recommendation algorithm based on term frequency inverse document frequency outperforms state-of-the-art recommendation algorithms on the recommendation accuracy.(2)Analyzing the problem of privacy disclosure caused by point-of-interest recommendation in geosocial networks,a privacy protection algorithm is proposed based on k-coordinate mean.According to user privacy requirements,the anonymous area is determined.The average coordinates are computed by at least k users' locations in the anonymous areas.The users' real coordinates are replaced by the average coordinates.The optimal anonymous area is selected to improve service quality.Theoretical analysis and experimental results demonstrate that the proposed algorithm reduces the anonymity time and cost,and applies it to the recommendation application without affecting the service quality.(3)A frequent location privacy protection algorithm is proposed to solve the problem of user privacy leakage caused by point-of-interest.According to the constraints set by the system,a frequent location set is constructed for each user.Due to the different background knowledge,hyperedges are composed of by subsets of frequent locations.Taking bias of user and location as the optimization goal,some hyperedges are remerged.The experimental results indicate that the frequent location privacy protection algorithm is compared with existing related algorithms which can effectively improve the problem of frequent location privacy which reduces bias of user and location.