Community Detection Based On Location Prediction In Social Network

Nowadays,social networks have permeated all aspects of people's life,and the research of social networks has always been the focus of attention of scholars.With the development of social networks,online activities are becoming more and more unable to meet the needs of people,and the offline activities are becoming more and more popular.In order to improve the success rate of offline activities,the organizers need to recommend targeted activities to the users in the neighborhood of the activity.However,with the enhancement of people's awareness of the protection of self-privacy,it is difficult to get the complete location information of users from the social network.In view of the fact that the original user location prediction does not consider user participation in offline activities,and user location factors are not considered in community discovery,a more accurate location prediction algorithm and a more effective community detection algorithm are proposed on the basis of user location prediction and community discovery algorithm at home and abroad in this thesis.For the problem of user location prediction for community discovery,after in-depth research and analysis of the label propagation algorithm,we found that the phenomenon of position label "countercurrent" will appear during the iterative process of the algorithm,and the update of the node location label has the problem of randomness,and so on,for solve these problem,we propose a location prediction algorithm based on label propagation algorithm,which is Label Propagation Algorithm-Location Prediction(LPA-LP).First,we calculate the k-jumps the public neighbor of any two points in a social network diagram,and select nodes with maximum similarity and their k-hop neighbors as the initial sets to do label propagation,then we calculate the degrees of nodes that aren't in these sets for these sets,and during every iteration,the node adopts the asynchronous update strategy and selects the largest node to update the location label.These operations can avoid the "countercurrent" phenomenon of position label and reduce the problem of randomness to update the position label,therefore to improve the efficiency and accuracy of the algorithm.The experimental results show that the improved algorithm improves the accuracy of location prediction and reduces the time cost of the algorithm.Aiming at the problem of community discovery based on user location prediction and different view of the background,this thesis proposes an algorithm based on user location prediction,which is Community Detection Based on Location(CDBL).In this thesis,at first a preliminary classification is made according to the location of the offline activities,then the breadth-first algorithm of graph is used to predict the user's position probability as a parameter of the objective function discovered by the community,based on the aggregation of user neighbors based on the offline activity locations.And set the threshold of community density as the lowest value of the output community sub-map,as long as the community sub-map that meets the predetermined threshold of the community is output as the community defined in this article.The nature of the communities found is that users in the same community are not only physically close to the social network but also physically close to each other.Experiments show that the proposed algorithm is not only suitable for dealing with large-scale social networks,but also can find a defined community from social networks.Experiments show that the proposed algorithm has high accuracy and stability,but also proves that the algorithm can be applied to large-scale data sets.