Analysis The Role Of Complex Network Nodes Based On Community Detection

In the real world, many large systems can be viewed as a complex network system, a lot of network systems have internal community structure. The node analysis is the core element in complex network analysis, the research of complex network in the past, often analysis the node by analyzing the importance of node in the network, and many scholars have proposed different indicators of the importance according to the overall characteristics of network. Meanwhile, community detection as a basic tool of complex network research has made great progress in the past decade, especially those based on the module algorithm and its modification has been widely used. In this thesis the method of community detection and the method of node importance analysis are combined to be used to analysis the node role in complex network, in this way we can find the important node accurately.Because of the existence of community structure in complex networks, the traditional methods of node role analysis can not find those important nodes, which only much important than other nodes in a community. Aiming at this problem, this thesis analysis node role after using the community detection algorithm to complex network. Moreover, aiming at the problem of low efficiency and locally optimal solution of module algorithms,this thesis presents a module-based algorithm by merge community label and make verification on the simulated data, the results show that the algorithm’s running efficiency and accuracy have improved. At last, this thesis constructs a real complex network by crawling the real social network-zhihu, and analysis the node roles of users, then summed up a multi-layered community detection method.The main work of this thesis as following.(1) Proposing a node role analysis method by community detection, for the real social network-zhihu, this thesis program a crawler to get the users and topics information, analyze the characteristics of these data, and build a topic communication network, then infer the mutual influence relations of social network users by using the community detection algorithm in the topic-propagation network to analysis the role of users.(2) Proposing a module-based algorithm by merging community label. In this algorithm, using community merger instead of the original node merge in module algorithm in order to improve operational efficiency; while introducing clustering coefficient to optimize local optima and improve the accuracy of the result.(3) Implementing a community discovery algorithm simulation platform, which integrates different community detection algorithms and evaluation methods. In this platform, some simulation data such as small-world networks and scale-free networks can be generated, and researchers can analysis the running state of algorithms in different networks; at the same time, this platform can load real data with fixed format for analysis.