Clustering Analysis Of Complex Networks Based On Computational Social Science

The study of complex network theory could help human to understand the complextopological relations between the objects and the dynamic behavior. And it has played animportant role in the internet, sociology, biology and other research areas. Clustering analysisis an important aspect of complex network research, and it is meaningful in many fields, suchas, understanding the network topology and functional characteristics, mining the potentialsignificance of communities, information recommending and the forecasting of behavior.Previous studies of communities in the real system, only research the model structureinformation, so they ignore characteristics and significance of the objects. The emergingdiscipline-the computational social science, which is related to the social network analysis,political forecasting, computational science and complexity science, aims to collect andanalyze data to reveal patterns of individual and group behaviors. Thus, it could help thesociological research from qualitative analysis to quantitative analysis. Therefore, this thesisapplied the research method of the computational social science to guide the communitydiscovery process in the large-scale complex network. It not only can make up for thedeficiency of the priori knowledge in complex network, but also can help to find meaningfulcommunities, meanwhile it can promote the verification of social theory in large-scalenetwork.First, this thesis makes a simple analysis for node importance index, power-lawdistribution characteristics, the weak ties theory. Second, this thesis analyzes the threeclassical community evaluation criterions of weighted network based on the heterogeneousweighted complex network simulate dataset, and adopts four datasets to compare theperformance of three typical clustering algorithms. Third, in combination with the power-lawdistribution theory, the weak ties theory, two-step flow hypothesis and so on, this thesis givesa new complex network clustering algorithm based on computational social science, andcompares the new algorithm with the three clustering algorithm combining with four datasets.Finally, based on the complex network control research model put forward by Liu andBarabasi, this thesis gives a preliminary study for the modified model based on the idea ofpropagation immunization and community structure, and gives a preliminary exploration of complex networks controllability based on the community structure, and the research bringscommunity a new criterion.