A Social Tie Inference Method Based On False Location Identification

With the development of mobile internet and the popularity of smart mobile devices,location-based services and social networks have been widely used.In recent years,Location-Based Social Networks(LBSNs)have received extensive attention from both industry and academia.Location-based social network is a network generated by adding location information obtained by GPS and Wi Fi,which is of great significance for understanding human behavior and e-commerce.Proximity of physical locations and mutual friends in a specific location provide valuable information for social network analysis.However,wrong location information can reduce the accuracy of social relationship inference.In order to solve this problem,this theses studies the identification of false location information in LBSN,taking high-quality acquisition of location information of social network users as an entry point,and designs a social relationship reasoning method and computing framework based on false location identification.The main works include:(1)Considering the impact of false location information on accurately inferring social ties,we propose a False Location Identification Algorithm(FLIA).Based on the combination of GPS accuracy and Wi Fi signal strength,busy public places are marked as false locations by comparing with predetermined thresholds,in order to more accurately represent the location information of LBSNs.(2)The social ties inference framework with two kinds of social ties inference methods based on LBSNs is formulated.The first method uses existing Social Tie Inference Framework(STIF)to identify social ties between different users;the second method can identify and eliminate false location information by using FLIA,thereby more accurately representing social networks,it is a more effective social tie inference method.(3)The performance of the proposed algorithm and improved framework is evaluated using three datasets: CDR,Brightkite,and Gowalla.The LBSN dataset was compared with the traditional computing framework STIF and the social tie inference framework applying FLIA.The experimental results show that when using FLIA,the social tie inference framework performs better than STIF in terms of precision,recall,and F1 value.This proves the effectiveness of the proposed FLIA,that is,deleting the location information of busy public places can effectively improve the social tie inference performance of LBSNs.