The Research Of Community Discovery Algorithm In Social Network

Social network is one technology involved in the field of cross-over study ofmobile technology, Internet technology, sociology research disciplines. The mainresearch interests include network formation model, the study of community found, thestudy of information delivery, security model studies. Social network research hasfocused on the nodes of social property and social relations, Studies have shown thatthere is a strong community structure in the real network. Community discovery is tofind community relations from the logical link between the overall network nodes.Other technology research is often based on the result of social network communityfound. So community found is the foundation of social network research.In this paper, we analyze and study society found deeply. Our work is mainlyfinished in three aspects: first, study and analyze the performance of existing classiccommunity discovery algorithm. Then, design a label based community discoveryalgorithm, and simulate it in simulation platform, which is accomplished by myself withthe general community discovery algorithm. Compare the algorithm with classicalgorithms under simulation and proof algorithm performance. Finally, we design anexploratory distributed community discovery method.This paper we describe the basic situation of the current community discovery study.We describe the classic community discovery algorithm and its development, throughthe analysis of all kinds of classification of community discovery algorithm. And furtheranalyze community discovery algorithm based on labels. Label based algorithm,although does not have a precise mathematical model, but became one of the most rapidcommunity discovery algorithm, because of its simple and efficient performance in thetest, which is suitable for large networks. The label based algorithm is more unstable asa result of too much random strategy in the algorithm.Through analysis advantage and existing problems of the current label basedcommunity discovery algorithm, we propose an improved community discoveryalgorithm based on tags, simple initialization and global label interactive communitydiscovery algorithm (SGLPA). The algorithm uses initialization algorithm from the node label and improvement of label update policy as the main idea. Thus, reduce theinitialized tag numbers of association found and numbers of random strategy in theprocess of updating. Then improve the computing performance of the algorithm and thestability of the algorithm. This paper we design and implement a general communitydiscovery algorithm simulation platform. We realize the classic corporate divisionmethods and the algorithm designed in this paper in the simulation platform. Finally,analyze and compare the simulation results.Last, we exploratory design a new kind of distributed community found methodbased on node label interaction, using centralized community discovery algorithm fordistributed computing.