Analysis And Forecast For User Mobile Behavior Based On The Check-In Data

With the advent of the era of Web2.0, more and more network applications have been developed. Among them, Social Network Services (SNS) is the most attractive one. From Facebook, Twitter to Renren, Kaixin, SNS has become an application with the largest user amount, the widest spread range and the highest commercial value. Recently, with the popularity of mobile devices and the rise of the location-based service, traditional social network and positioning technology integrates to generate a new kind of online social media-Location Based Social Networks (LBSNs).LBSN is the combination of position and social. It allows users to share their location information with no limitation. In LBSN, the most important unit is user and position, which could generate check-in data. Personal check-in data could represent a user’s history trajectory. A lot of user check-in data could reflect the human movement pattern and the law of life. The check-in data is a kind of social data with geographical information. It can not only reflect user’s social behavior, but also reflect user’s mobile behavior. At the same time, because of its simple acquisition mode and low cost, more and more researchers take advantage of check-in data to study human movement in recent years.In this paper, the location-based social network data has been studied from three different aspects, which is spatial, temporal and social, to mine the characteristics of user mobility. Through a series of data pre-processing and statistical methods, some impact factors have been found and have been used to establish a location prediction model. Based on this basis, the thesis further proposes a framework of the location prediction system. Finally, this paper also explores the application research of location prediction-the group recommendation. Location prediction can detect users within a certain range. Recommending content to these users can effectively reduce the burden and the cost of the network. Within this framework, this paper proposes a group recommendation algorithm which integrates group members’ personalized recommendation results, and then realizes the simulation of group recommendation scenario and performance evaluation by combining the LBSN data and Movielens data. At last, this paper proves the feasibility and effectiveness of the application research.At the end, based on the theoretical analysis and the experimental results, the improvement and research directions are proposed for future study.