Personalized Point Of Interest Recommendation Algorithm In Location-based Social Networks

With the rapid development of web 2.0 and positioning technology,users can easily get real-time location,and also led to many location-based social networks.A good recommendation system does not require users to explicitly express their needs,but to analyze the user's historical behavior through modeling and automatically help users find information they are interested in.In recent years,the application of recommendation technology to location-based social network has attracted the attention of scholars at home and abroad,and has gradually become a popular research direction.This paper deeply studies location-based social network and point of interest recommendation algorithm.For the users' interest varies with time and space,this paper proposes a point of interest recommendation algorithm incorporating temporal-spatial and popularity influences.Firstly,it combines user-based collaborative filtering with temporal influence and gives a corresponding estimation method.Secondly,it combines time-based point of interest popularity with spatial influence and gives another corresponding estimation method.Finally,it gets a point of interest recommendation algorithm based on the linear framework by combining the two methods.In addition,for the problem concerning sparse user-interest matrix,this paper proposes a point of interest recommendation algorithm based on latent semantic analysis model.The algorithm reduces users' check-in matrix dimension to get user matrix and POI matrix by latent factor model.On the other hand,this paper mines users' preferences from users' review texts by probabilistic topic models LDA.Lastly,by fusion of various heterogeneous information such as time,space,popularity,history,and review text,this paper puts forward a personalized point of interest recommendation algorithm in location-based social networks.On the real Foursquare and Yelp data sets,this paper verifies the performance of the proposed algorithm.Experimental results show that,the proposed algorithm has better recommendation results at accuracy and recall rate compared with the current same type of algorithm.