Community Detection Algorithm Based On Similarity Index

Complex networks have the characteristics of self-organization,self-similarity,small-world effect,scale-free,and community structure.Community structure is an important attribute of complex networks.Analyzing the community structure is very meaningful.It is equivalent to the functional units of the network,such as a cycle or path in the metabolic network,or corresponds to a collection of topics on a web page.In addition,some recent results have shown that the attributes of the network at the community level are completely d ifferent from those of the entire network,and ignoring the community structure may miss many interesting features.In the past few years,a large number of methods have been developed to detect the community structure in the network.For a particular network,what kind of calculation or measure is most effective depends not only on the specific problem application but also on some of the behavioral characteristics and characteristics of the system itself.So each method has the adaptability as well as the advantages and disadvantages of each method,but in general it has its own scope of application,and it has practical significance and value for some actual networks.Based on cosine similarity and Euc lidean distance similarity,the paper proposes a mixed similarity index,which can accurately measure the degree of similarity between multi-dimensional vectors,and uses this index in multi-attribute large group decision making problems.A multi-attribute large group decision making algorithm based on this index is proposed,the algorithm can help managers to better make decisions on complex decision-making problems.The multi-attribute large group decision-making system is also a complex system.Abstracting the system,a complex network can be obtained,and then the indicator can be extended to the complex network research field.Based on the mixed similarity index,a community detection algorithm based on the index is proposed.The algorithm initially treats each node as an independent community structure.The idea of hierarchical clustering first merges the most similar nodes and then Community blocks are merged according to community similarity,until the modularity is maximized and the algorithm terminates.Experiments on real networks such as computer-generated networks,karate club networks,bottlenose dolphin networks,and soccer networks,proved that the feasibility and advantages of the algorithm are proved.It shows that the algorithm not only can correctly divide the community structure in the entir e network,but also improve the accuracy of community classification.Finally,using the journal data on “complex networks” from 2013 to 2017 in the “China National Knowledge Infrastructure(CNKI)”,the paper constructed a keyword network with time series.Using the above algorithm to study the evolution of the community structure in the network and reveal the law.