| With the development and popularization of the Internet, online social networks havepenetrated into every corner of people’s lives, people closer to the distance between each other,the study of social networks allows us to understand the structural characteristics and evolutionof social networks to let it better serve humanity. In the research of networks has two hot issuesthat has a very important meaning for our understanding of the complex social structure andthe characteristics of the network,that is community detection and evaluation of nodes.Community detection work can split large network into smaller communities granularity, Letus find groups that Individual contact closely.Node assessment can assess the importance ofnodes in the network from different angles, find important nodes.Most existing community detection algorithm base on graphic segmentation andhierarchical clustering ideas, although in most cases these algorithms can effectively identifythe community, but the community must specify the number or size of the community, it isclearly unreasonable. Genetic algorithms as a method of searching the optimal solution canautomatically identify a priori information in the case of the absence of the number ofcommunities, efficient and accurate discovery of the community, but the traditional method ofinitializing the population is only taken into account adjacency information, did not fullyconsider topology of the network.Therefore, the quality of the population is poor, and impactof the convergence rate. There are many nodes evaluation algorithms,some based on localfeatures, some based on the topology of the entire network. PageRank algorithm as analgorithm to assess the importance of web pages in search engines field， also has a very widerange of applications in assessment of the social network node． However, in the traditionalPageRank algorithm the weights are evenly distributed，this kind of distribution in a socialnetwork is unreasonable, because the social network reflects the relationship between users, thecloseness of this relationship is points can not be treated equally. For the above analysis, Thispaper focuses on improving the fault that existing on genetic algorithms for communitydiscovery and PageRank algorithm for node assessment the main work is as follows: First, the population initialization method of the genetic algorithm used to detectcommunity has been improved, according to the characteristics of the network, given thecharacteristics definition of the dissemination of information in the network, Then proposed tofully use the network topology initialization method k-path method according toself-characteristics and dissemination of information in the network characteristics of theirsocial networks, And gives the genetic algorithm based on k-path initialization.Then, based on the closeness between nodes, put forward the concept of recognitionbetween nodes. For the problem that it is unreasonable evenly distributed weights when use thepagerank algorithm to assess the node,proposed that the recognition between nodes as the basisfor allocation of weights. Finally, given the improved node evaluation algorithm ARankAlgorithm.Finally,we verify the improved genetic algorithm and PageRank algorithm in the dataset.The experimental results show that the improved genetic algorithm for community detectionfaster than the traditional algorithm in the convergence rate, the results of assessment of nodesthat get from the improved PageRank algorithm are more reasonable than traditional methods. |
