The Research And Implementation Of Location Recommendation Algorithms Based On Location-social Information And Groups

With the development of mobile Internet,smartphones and GPS services are becoming more and more popular,and location-based social networks(LBSN)are gradually being incorporated into people's daily lives,and users can share their visited location in time.Location-based social networks and the massive data help to develop the location recommender system,give users better service experience and help the location business to catch users' trend.Therefore,this paper focus on the research of location recommendation.In LBSN,it is obviously deficient to recommend locations by traditional recommendation algorithms,because user's behaviors are influenced by many complicated factors.It is necessary to study how to apply the information in the location social network,such as the social relationship,geographical information,user's bias.On the other hand,it is also necessary to ensure the recommendation efficiency in large data.All the above are difficult to solve by traditional recommendation algorithms.The main contributions of this paper are as follows:(1)this paper proposes a recommendation algorithm based on social relationship and trusted-behaviors,which considers the social intimacy,trust,and power law for distance;(2)this paper proposes a location recommendation based on spatial-temporal information and latent topics,which considers the categories of location visited by users,the spatial-temporal preference of users,and probabilistic topic model;(3)this paper proposes a location recommendation algorithm based on latent groups discovered by clustering according to different information like social relationship,interest and geographic location,which help to improve the speed of recommendation;4)this paper gives a prototype design of location recommender system,which combined the algorithms above and can give recommendation with different strategies according to users' status.Experiments on the datasets from the Gowalla and Foursquare shows that the algorithms proposed outperforms other algorithms in terms of some metrics,such as precision,recall and time cost.