Community Spatial And Evolution Analysis In Complex Networks

Community discovery and its related analysis have been paid much attention during the exploration process (a community is a kind of special structures in the complex networks). Researchers devote themselves to proposing many approaches to detect communities from networks in the past several years but those methods just focus on flat topology analysis. As to how to analyze community attributes further, especially how to mine out spatial and temporal information of communities, is starting up. By getting spatial information, communities are displayed in a 3D view and by getting temporal information, development or evolution traces of communities are obtained to make their developments observed. Therefore, this paper concentrates on these two points: revealing community spatial structure and tracking community evolution.Unlike many existing methods to reveal spatial structure of a community (they just extract sub-communities locating in the low level from communities in the high one), the method in the paper is content-oriented, that is, it is based on the difference of community contents to distinguish community levels: which community has high or low social status. The processing steps are as follows: (1) determine node activity scope; (2) analyze the scope quantitatively; (3) confirm node hierarchy values; (4) discover communities; (5) With the combination of node hierarchy values and communities, the spatial structure of communities is ultimately revealed. In the end, people get a new view to observe social networks.With the goal to reveal how the state of a community changes in the next snapshot, the method tracking community evolution in this paper effectively utilizes core nodes in a community to establish evolution relationships. It involves two steps: (1) find core nodes; (2) build evolution relationships based on core nodes. Using those unrevealed traces, the development trend of a community is able to be predicted. Compared with other methods, my method is parameter-free and has capability to discover split and mergence points.The experimental datasets include: co-authorship networks, call networks, the email network of Enron, movie actor collaboration networks, Internet, vocabulary networks, software networks and so on, through which those methods given in the paper are validated. In the experiments of spatial analysis, the position status of Enron communities is uncovered: three communities are in the leading level and six in the ordinary employee level. In the experiments of temporal analysis, under diverse datasets, we reveal evolution traces and their associated characteristics from which feature difference in social networks and nonsocial ones are analyzed (For example, in social networks, a community with lower stability will live a longer life and it is opposite in nonsocial ones). According to those differences, a novel method to tell social networks from non-social ones is provided, which is validated to be correct in our datasets. Moreover, such differences are also useful in the improvement of building social network models.