Community Discovery Based On Multi-subnet Composite Complex Networks

Community detection is a popular problem in network, the existing algorithms is based on the problem of community detection based on the network with an kind of node and a kind of relationship. However, the existing network is always consisted of with different kinds of nodes and different kinds of relationships, we call this kind of network as multi-subnet composited complex network or composite network for short,there are traditional network but also cross-subnet community, and it is always meaningful, for example, phenotypic-protein complex network, studying it is is useful to find the pathogenic gene. Many problems can be treated as community detection in network, but the uneven linking density in the network with different kinds of nodes and different kinds of relationships lets the existing algorithms can't find the cross-subnet community.This paper studies the current simple network and multiple subnets association discovery algorithms, and analyzes the advantages and disadvantages of various algorithms.This paper proposes a algorithm for the network with uneven linking density and different kinds of node.In the light of Newman test network, this paper structures the test network with uneven linking density and different kinds of node. The experimental result shows this algorithm successfully finds the community structure in the composite network with uneven linking density and different kinds of node. This paper structures phenotypic-protein complex network, and uses the discovery algorithm to find the community, it is useful to find the pathogenic gene.The main research in this paper is as follows:(1)To solve the problem of the uneven linking density, based on the idea of the information dissemination. This paper proposes a algorithm that nodes are expressed as vectors by the edges between subnets and the vectors of the node in the network is done by zero-average standardization, the community will be found by clustering the vectors.(2)In the light of Newman test network, this paper structures the test network with uneven linking density and different kinds of node. The experimental result shows this algorithm successfully finds the cross-subnet community structure in the composite network with uneven linking density and different kinds of node and it overcomes the the problem of the density in the network.(3)This paper structures phenotypic-protein complex network with two kinds of nodes?three kinds of relationship?17714 nodes and 1426 edges between subnets, and uses the discovery algorithm to find the community.