Friendship Prediction In Social Networks Based On Location Information

The prevalence of Social Networks has changed the way people communicate, more and more people like to socialize and share kinds of information in it. With the development and popularization of positioning technology, Location-Based Social Networks are also becoming more popular. How to use vast amounts of user data for location-based social networks to analyze user behavior, predict the friendship between users becomes an important issue.In this paper, we analyze user behavior, predict the friendship between users based on a large number of user data for Location Check-in Service. The main work of this paper includes the following four parts:(1) Using users’ check-in record, paper analyze users’ check-in behavior from the three aspects of check-in frequency, check-in location and check-in time, found that the number of user’s check-in record and the number of users following the power-law distribution, and analysis the location and time that users check-in more concentrated;(2) From the three aspects of moving distance, radius of the active area and moving periodicity, we analysis the user’s mobile trajectory through the locaton information of user’s check-in record. From the study of moving distance, we found that user’s mobile trajectory to travel in real life is matching Levy Flight model; from the study of radius of the active area, we found that the radius of active area for each user and the number of users follow the power-law distribution, indicating that most users travel locally, with few long-distance checkins; from the study of moving periodicity, we found that users will periodically appear in certain locations, according with the law in real life;(3) Through experiments, we found that there may be more friendship between users with a common check-in behavior(co-location). In this paper, we further explored the relationship between users’co-location and friendship from the three aspects of the number of co-location areas, the number of co-location and the temporal range of co-location;(4) Based on the previous, this paper improves the way geographic area divided by location entropy to distinguish the difference between different geographical areas, and then proposed a variable scale division method of the geography. By experimental verification, when using uses’ co-location features to predict the friendship between users, the variable scale division method of the geography becomes more effective.