Exploring Location Perdition In Mobile Social Network

With the thriving of Internet, social networks have become a popular form of social media recently, while Social networks and location-based services began to coalesce along the increasing popularity of mobile terminals and positioning technology. Then, mobile social networks come into being. They provide location related services that allow users to “check-in” at geographical locations and share such experiences with their friends. Large numbers of “check-in” records continue to accumulate over time, which contain rich information of social and geographical context and provide a unique opportunity for researchers to study users' human mobility pattern, which in turn enables a variety of services including place advertisement, intelligent transportation, and Urban Computing.In this paper, we mainly study the problem of user's location prediction by mining based on the “check-in” data, and the contributions of our work are summarized below:1. We put forward a prediction algorithm based on the temporal features. The prediction algorithm predicts users' mobility based on Markov model, and then combines the temporal features of to amend the prediction results. By evaluating our algorithm on two real check-in dataset, we find the prediction performance of our algorithm enhance 15%~20% compared to competing baselines.2. We study the existing prediction algorithms that mainly relying on regular mobility patterns, and find this kind of algorithm can only make a prediction on the visted location, but they can not describe users' check-in preference on the unvisited locations. Hence, we proposed a location prediction algorithm combines with collaborative filtering. We not only propose a set of features that aim to describe the regular mobility patterns, but also make full use of the collaborative social knowledge to analysis the users' mobility patterns on the unvisited locations. Finally, we further extend our study combining all individual features in a supervised learning models, based on M5 model trees and linear regression, to get a higher overall prediction accuracy. By evaluating our algorithm on two real check-in dataset, we find the prediction performance of our algorithm enhance 19%~21% compared to competing baselines.