Research On Algorithms For Community Detection In Social Networks

With the rapid development of information era and information industry, moreand more people are living and working is closely connected with the network, thenetwork become one of the tools for people to live. It not only contains personalinformation, but also contains the group information, this makes the study of complexnetworks has become a research hotspot at home and abroad. This paper made someresearches of community classification threshold selection problem and the similaritymeasure between the nodes based on complex network. Main work is as follows:(1)In terms of community classification threshold selection, based on existingmodule through theoretical analysis and actual measure are derived, points out itsdeficiencies, then based on the fuzzy clustering theory put forward a new communitydivide index (FF-FT) which is based on the “perfect networks” and edge densitytheory. Through theoretical analysis and experiments show that it not only overcomesthe original module’s defect which can’t division communities accurately, but also hasmore advantages than other module in the community division degree.(2) Considering that in a complex network, the global similarity standard is ofhigh complexity and there is a defect that local similarity standard has biased error onnode similarity the paper puts forward a community discovering algorithm which isbased on the degree of contribution of shared nodes in complex network. It distributesthe contribution, using the degree of shared nodes as the unit, equally which is madeby shared nodes similarity more accurate. The network experiments of PPI net workand USAir and HSM standard, ERS standard has a more veracious forecast to thesimilar relationship between nodes. The experimental results of artificial network andtypical complex network show that ERS algorithm has a better stability and couldadapt better to complex network than GN、FN、N-cut algorithm.