Research Of The Personalized Recommendation Technology On Location Based Social Networks

With the development of positioning technology, getting real-time location data becomes easy, Location Based Social Networks develop rapidly, more and more mobile terminal access to the internet to provide location information abundantly. Personalized recommendation system on LBSN, e-commerce and 020 are closely together, created a huge wealth, the research of personalized recommendation technology on LBSN has already became a hot area in academia sector today.This thesis explored the personalized recommendation system theory on LBSN, introduced the related technology and algorithms, study current mainstream personalized recommendation framework. This thesis also found out disadvantage of clustering technology and the friends recommend model, putted forward two new algorithms. One is a kind of clustering algorithm based on LBSN, for the purpose of getting better clustering results. Another one infused friends recommend model symmetry into the recommendation results, which can obtain better precision and recall. Based on this algorithm, the completely symmetry friend recommended result has better precision and recall.Firstly, this thesis introduced the throretical research background and significance, showed the related research work on LBSN and personalized recommendation technology, gaved the main work and the organization structure arrangement. This thesis mainly introduced personalized recommendation technology on theoretical foundation, contained the advantages and disadvantages of some recommendation algorithms. This thesis also showed the clustering algorithm on LBSN, analyzed the advantages and disadvantages of the application of the traditional clustering algorithm on LBSN, for the propose of getting a better distance sum of squares and convergence speed. This thesis analyzed the traditional TOP-n algorithm, which is not considering the symmetry between friends, putted forward a bidirectional TOP-n algorithm, this algorithm contained two kinds of incremental strategy for the construction of the similarity graph, transformed interests of user into directed graph, proved that this algorithm has a better precision and recall. The last chapter contained eight experiments, the results showed that the clustering algorithm in this thesis has better distance sum of squares and convergence speed. In addition, by comparing the sum of the square of distance between objects in the same cluster among different algorithms, the proposed algorithm is evaluated to obtain better clustering results and convergence speed over location based social networks. Comparing to the traditional k-medoids algorithm,the cost has obviously reduced, as for and the degraded k-medoids algorithm, the cost can be reduced among 1.2% and 2%. The bidirectional TOP-n algorithm has a better precision and recall. With the different number of recommended N and increment number k, the bidirectional TOP-n algorithm has better performance. Complete symmetry TOP-n algorithm has a recommended result which with less recommended number n and more related recommendation items, if this algorithm has the same recommended number n comparing with the traditional TOP-n algorithm, it has a higher precision and recall.