Research On Location Recommendation Integrating Trust And Personal Preference

Recently, with the mature of mobile positioning technology and the popularity of location-based social network (LBSN), a variety of location-based social networking applications emerged. Because the relevant location recommendation technologies provide us a new platform to understand users' behaviors and preferences based on their location histories in LBSN, it received extensive attention. Location recommendation technologies are used widely in Travel Route Recommendation, Location Navigation, etc. However, existing location recommendation technologies cannot extract the users'preferences accurately, and they do not consider the trust relationship between users. Thus, the recommendation is so bad that users' satisfaction is low. Therefore, users desire a novel method for the location recommendation based on LBSN.In order to achieve highly accurate recommendation, this thesis has designed a special users' preference storage structure, WCH-Set, combining the traditional recommendation theory of recommendation and the characteristics of LBSN. The main contributions of this thesis are summarized as follows:First, it proposes a data model based on the characteristics of LBSN. The users'data is classified according to their cities. Locations are designed as two-level category label according to their functions.Second, it proposes three off-line algorithms to extract the users'background, users'trust and the users' preference. It employs WCH-Set and makes extracting users'personal preference accurately possible.Third, Candidate Experts Selection algorithm is proposed to select the needed special-purpose users based on users'personal preferences and trust relationships between users. It not only meets the time requirement, but also supports Top-k recommendation.Finally, the online recommendation method based on traditional user-based recommendation theory is proposed. The algorithm supports WCH-Set similarity calculation.The theoretical analysis and experimental evaluations show that the methods for location recommendation are feasible and accurate.