| With the wide application of GPS technology and the rapid development of new generation information technology,Location-based Social Network(LBSN)has became an important direction of Internet application.Due to the users' social needs and the rapid development of cities,the LBSN site has accumulated a large amount of data.Point-of-Interest(POI)recommendation based on LBSN platform which integrating location-based services and social services has became a focus of personalized recommendation research.The point of personalized recommendation is to get user's preferences more accurately.Compared with traditional personalized recommendation,LBSN data contains more context information,which provides a new cut-in point for mining user's preferences.This paper utilizes the time information in the LBSN check-in records to discover the schedules of the user's check-in behavior,so as to satisfy the user's needs at the current time as far as possible.At the same time,we redefine the similarity users of the target users by using the trust relationship which been studied from friend relationship in LBSN.When measuring the similarity of interest points,the longitude and latitude information in LBSN is used to recommend the nearer interest points to users.The research work of this paper mainly includes the following aspects:(1)Analyzing user's check-in records from three dimensions: time,geographical location and social relationship.constructing a heterogeneous LBSN network which include user nodes and POI nodes,as well as three types of edges of user-poi,poi-poi and user-user.At the same time,this paper presents a weight measurement method for this three types of edges in heterogeneous networks.(2)Introducing meta-path to characterize the type of relationship between nodes,and calculating the relevance of nodes by using the characteristic value of meta-path,the algorithm enriches the effective data of target users by considering multiple meta-paths,thus alleviating the problem of extreme sparse data in LBSN.(3)Introducing metapath-based recommendation into group recommendation.And proposing a hybrid preference fusion strategy which takes into account the weight of group members and the difference of group preferences.The effectiveness of this strategy is verified by comparing experiments with other fusion strategies. |
