Location Prediction Based On User Behavior Analysis In LBSN

In recent years, with the rapid development of mobile terminal locating technology and mobile Internet technology, location-based social network(LBSN) platform has achieved great success. Location feature in LBSN has built bridges between the virtual social space and the real behavior space. It could fuse the online relationship with the of fline behavior. It offers new opportunity to track more fundamental law behind behavior of groups and individuals through interactive relationships and trajectories. LBSN platform provide a large number of data for the study of location prediction,at the same time, good location prediction algorithm also provides users in LBSN with convenience,and can create huge social and economic benefits.Currently, researches on user location prediction mainly focus on the next location. Real-time algorithms are proposed to predict the next location of the current one. However, those algorithms are quite limited in practical application as they cannot deal with the locations in the far future. Considering this, we propose a mechanism which could predict user location at any given time in the future.We first analyze the user check-in in the aspect of periodicity and geographic distribution, as well as the user social relationship. Then the factors which might affect user future check-in are explored in the view of location category and location itself. Based on this, we model and further quantify the multi-dimensional features related to the user check-in behavior. At last, we propose the location prediction algorithm based on multidimensional mixture features (LPMMF). With the algorithm, a prototype system of location prediction based on user behavior in LBSN is designed and implemented.In order to present the performance of LPMMF and analyze the effectiveness of each feature in it, we carry out the comparison experiment with the Foursquare dataset. The results indicate that features introduced in this thesis are effective, among which the check-in periodicity and user preference for venue are stronger predictors. The experiment shows the accuracy of LPMMF and proves it outperforms the baseline algorithm.