Community Detection Of Attributed Networks Via Nonnegative Matrix Factorization

So far, we have entered the complicated Internet time,there are variety of complex net-works in life, such as mobile phone communication network, transportation network, electric-ity network, and so on.Today,our lives and production activities depend on the safe,effective and reliable operation of these complex network systems .more importantly, many studies have shown that different complex network systems have the common concepts, methods and theory,which makes it necessary.The Network Science has been a new research field, and has rapid-ly developed.At present,the Network Science includes researching network property ,building network model,analysing network behavior,designing network performance. Community struc-ture is one of the common character of the many actual network, that is, the whole network is made up of some communities and the connections between nodes within each communities are closed together, between different communities which is sparser.Therefore, it is necessary to research the quantitative characterization of the community and the effective mining algorith-m of complex network structures.In recent years, community structure detection technology is significance in data mining, which has widespread applications in Sociology, Biology, Market-ing,created immeasurable economic benefits.In this paper,we use network edge information and the node features to build communi-ty detection model.Based on non-negative matrix decomposition and the covariant information,build the community discovery model.In this text, the node covariant information can be flex-ibly added to the new model.That is to say, the covariant is added to the model with different weights,due to the different affection of the community structure.Furthermore,the model can flexibly choose the effective covariant and eliminate the invalid covariant.According to this pa-per, we can see that the non-negative matrix factorization method with covariant has a very good simulation result and obtained the good effect in the application.