The Research Of Community Detecting In Signed Networks

In recent years, the research of complex network is a hot issue. Increasing numbers ofresearchers have pay attention to the area. A great deal of scientific research proved that manycomplex networks have a common property. The property is community structure. How to findthe community structure from a real complex network is a hot and important topic. Andresearchers have acquired some achievement. Almost of them just suit for unsigned network, inwhich there are only positive links. But, in the real world, many complex systems in the realworld take the form of signed social networks which contain positive and negative relations. Ifwe ignore the negative relations, we maybe miss some important property of the network.Community structure also an important aspect for the research of signed networks. The study ofcommunity detecting algorithms in signed networks has important theoretical and practicalsignificance. Community Detecting algorithms have been developed in the past. But, most ofthem are only effective for networks containing only positive relations and, are not suitable forsigned networks. Base on this aforesaid present situation, we propose a new algorithm to detectcommunities in signed networks. Both the signs and the weights of the links are taken intoaccount by the algorithm.In the paper, we introduce the definition of signed networks, properties of networks and thedefinition of community. Then we analyze modularity in the signed networks, and we presentanother form of the modularity which makes our algorithm understand easily. We introduce anew community detecting algorithm which can suit for signed networks. In the algorithm,modularity function is not just a evaluation criterion. It thinks about the function in theoperation process. In every step, the modularity have a largest incremental quantity. When weget a max value of modularity, we can get a good community structure. Both the signs and theweights of the links are taken into account by the algorithm.We use four data to test the algorithm by experimental method. Community structure by thealgorithm is same as the real community structure. And the value of modularity is high. So thealgorithm can work well. Then making a comparison of thisalgorithm and FEC in a computer-generated network, we know that the performance of ouralgorithm is superior to FEC algorithm. Then, we analyze a real data and a weighed complexnetwork by the algorithm, the results are consistent with the actual.The paper has two innovations. First, it proposes another form of modularity in signednetworks from the perspective of community. Second, it analyzes the proportions of commentnetwork based on the community structure.