Community Detection And Location Recommendation Based On LBSN

With the rapid development of intelligent mobile terminals and the rapid rise of social networks and location services,location-based social network(LBSN)come into being.At the same time,Information technology is developing quickly,all kinds of information which let us overwhelmed.In the face of serious information overload,it is difficult to quickly locate the information resources you want,personalized recommendation technology can quickly help users locate their favorite project,which can help users save more time in searching.In the location-based social network,the location recommendation is a new type research topic,but the sparseness of the user's check-in numbers seriously affected the recommended accuracy,the new users join and the new location join the system is also a large problem for recommended,and with the arrival of large big data,stand-alone computing is far from meeting the requirements of the current large amount of data.In this paper,the relevant issues related to the location recommendation are deeply discussed and researched,especially the following points.1.Based on the study of label propagation community detection algorithm,a Mapreduce label propagation algorithm based on global center point is designed.In this algorithm,the stability of community detection is considered.By comparing with the classical community detection algorithm,the stability of community structure is improved,and in the face of the large amount of data found in the community,this paper has designed the Mapreduce programming model implementing the algorithm,improving the scalability and computational efficiency of the data.2.Combining the community detection algorithm based on the global central point with the cooperative filtering recommendation algorithm.That is,on the basis of the detection of the community,taking the community factors into account and users in the same community have similar interests,which alleviate the cold start of new users.Also in the traditional collaborative filtering algorithm,the user in the community did not score the location,which will impact on the final user location score.In this paper,the user location score is combined with the number of user checkpoints to the location,and a method of filling the user's location score is proposed,which alleviates the sparseness of the data in the recommendation process.