Research On The Model Of Latent User And Location Recommendation In Location-Based Social Networks

Social network is organized with people in physical world. Location information is added toexisting online social networks and being part of user’s basic information. Online users could sharetheir present location information on a website, and mobile users could upload their location withthe smart devices. This type of social network combined with location information and socialrelations is called as location-based social network. The dimension of location brings socialnetworks back to reality, bridging the gap between the physical world and online social networkingservices.The thesis studies on this type of location-based social networks. Users could do check-in actionto say that they are currently there and share this information to their friends. And then their friendsmay check in at this place in the future if they are interested in the location. This type of check-inapplication aims for enriching user’s social experience.The thesis analyzes the background, development actualities and significance of research on thistype of location-based social networks. Check-ins collected from the global famous check-inwebsite Fourquare.com are experimented in this paper. The main work and contribution come out ofthe thesis are:(1)Mining the relation between location and user: study the impact of physical location andpeople’s check-in mobility. And present analysis of the character of location where people checkedin;(2)Mining the relation between locations: study the relation between user’s check-in places andalso the relation between user’s and their friends’location;(3)Research on recommendation model: both friend suggestion model and locationrecommendation model are presented in the paper. Graph model and probabilistic model are used tosolve the friend suggestion problem. As for location recommendation, such check-in sequenceshistories and the hidden topic information about locations are applied to train and construct themodel. The results of simulation experiments demonstrate a high accuracy and recall rate for bothuser and location recommendation problem.