Research And Application Of Recommendation Algorithm In Location Based Social Network

With the progress of wireless communication and location-acquisition technology,and mobile devices with positioning function become more and more popular,make it easy for people to introduce location service function in the traditional social network.The popularity of LBSN(location based social network), shows a new platform for us to understand users’ preferences and behavior based on their location histories. There are some problems in studies of recommending friends and locations for user, like the complexity of user’s location information and invalid data, user similarity calculation method of single, user-location matrix sparse bring recommenddation result is not very good,etc.The paper based on the "new media interactive broadcasting system" project, which is a social network platform based on media services, and have the characteristics of LBSN after bring the location attribute.As the background of this project for data preprocessing, data analysis, data mining, finally realizes the user’s friends and the locations of the recommendation.The main research contents of this paper are as follows:1.data preprocessing algorithm: which is aimed at the complexity of user’s location information and invalid data, preprocessing the original data, eliminate invalid data and outlier data.2.similarity measure: in order to enrich the similarity calculation method, proposed the concept of user interest degree, trust degree, preference degree. Trust degree which is based on intimacy of users in social relations; interest degree which is measure two users in the location of the interest similarity; preference degree which is according to the semantic information of users check in the past to learn the user’s personal preferences.3.friends recommenddation: according to the regional characteristics of the user data, combined with Geo-Location information and user relationship, comprehensive consideration the similarity of user interest and trust, proposed the interest and trust friend recommendation algorithm.4.location recommenddation: according to the cold start and user-location matrix sparse problem, proposed the combined services algorithm based on preference and interest and trust recommendation algorithm.The results of experiment show that, the method are recommended in this paper has better recommendation effect, and applies this method to the new media nteractive broadcasting project.