Research On Personalized Recommendation Method Based On Location Based Social Network

With the further development of Internet technology,social networking is also increasingly popular in people's lives up.People use mobile social networks to find useful life information,share life experiences,and interact with friends.And with around 2010,Andrews,Apple and other smart phones generally universal.Based on the original social network,we combine the information of user's geographical location and form a new social network concept: Location-based Social Network(LBSN).Location-based social network not only concerned about the information published on the user line and the user line of friends,but also retains the activities of the user line range.Understand the user's daily behavior.As LBSN contains a large number of types of information.We can mine the data and find meaningful information.At present,the use of LBSN data on the user's personalized recommendation of the research is very high heat.At present,the use of LBSN data on user recommendations are mainly concentrated in three areas:(1)points of interest recommendation: This recommendation is for the user,including two recomendations,one is the destination,the other is a combination of multiple locations Line recommendation;(2)business site recommendation: This recommendation mainly for businesses;(3)Friends recommend: the user online friends recommended.LBSN recommendation system has made great breakthroughs,but there are still some problems:(1)Points of interest and users are large,the amount of calculation is big,and data is sparse;(2)the user relationship representation method is not practical(3)the new cold-start problem of adding users(2)the problem of the user's cold start;(4)There is no contextual information.In view of the above problems,this paper presents some new ideas and solutions(1)Solve the problem of data sparsity by matrix decomposition and demographic knowledge;(2)to enrich the relationship that friends,more in line with the real world.In gene-ral LBSN data,the relationship between friends only 01(with or without)that is clearly not taken into account the closeness of friends,this paper will combine online and offline data,combined with the community found that the algorithm found the degree of interest between friends coincidence,To quantify the relationship between friends,so the recommendation system to play a more meaningful role.(3)In view of the problem of context information,this paper proposes a combination of time and space information,proposed the combination of the user's current time and space recommendations,analysis of users and points of view of time and space characteristics and the current time and space information to do better recommendations.In addition,this paper will combine the online information and offline information,taking into account the actual geographical some of the geographical attributes,proposed a new business location problem.And propose a friend-recommend model based on random walk between users and points of interest.In the experiment,we select the more famous LBSN abroad data Foursquare analysis and experiment,through the accuracy rate,recall rate and some other criteria to compare the text of the method and the previous research results.A series of tests on the algorithm of the article,to achieve the user and the POI supplier of effective recommendations.