The Research Of Motif-based Measurement And Community Detection Algorithm In Complex Networks

With the development of study on complex network, community structure characteristics are widely used in biology, physics, computer graphics and sociology fields. Finding community structure has very important practical significance. With the research of network topological properties, people find motif structure in complex network. Motif is one of the basic structures in the network, and its size is between a single node and community. We should consider the existence of motif when we study the properties of community structure.This paper studies the measurement and community structure division based on motif in complex network, mainly includes the following content.(1) We make motif detection analysis with Rand-ESU algorithm for nine networks of different sizes, and we verify the existence of motif structure in network. We make detailed motif analysis for Dolphin network, Social network and Science network, including the kinds of motif and structure characteristics. We put forward the definitions of node degree and edge degree based on motif, and we study the correlation between the new definition and the traditional definition. We verify the rationality of the proposed new definition by the correlation research.(2) We introduce the definitions of community structure and modularity based on motif, and we put forward improvement GN algorithm based on motif modularity. The Dolphin network is used for simulation analysis, and the superiority of improved GN algorithm is verified through the comparison between algorithm’s community division and real community division.(3) We put forward edge clustering coefficient definition based on motif, and we put forward community detection algorithm based on motif edge clustering coefficient. The stock network is used for empirical analysis. We study the motif structure of stock network, and we divide community of stock network with the algorithm based on motif edge clustering coefficient. We use deliberate attack and random attack strategy to attack the stock network, and we verify the stability of the algorithm by comparing diversity of the community division.This paper puts forward the measurement and community structure division based on motif in complex network. The experiment results show that the research of network measure theory and community structure theory based on motif is meaningful.