Research And Implementation Of Overlapping Community Detection Algorithm In Complex Networks

In the era of big data, with complex network data continuing explosive growth, the impact brought about by complex networks has entered every aspect of our lives. Therefore, it has an important value and big significance to research the detection of community structure in these complex network. Complex network community structure can be divided into non-overlapping and overlapping community structure, compared with the non-overlapping, overlapping community structure is closer to real community organizational structure of the network. However, research of overlapping community structure is currently in its infancy, but due to the high complexity of the overlapping community detection algorithm, so researchers need greater efforts. This thesis aims to explore the accurate and effective community detection methods for complex networks. We proposed one non-overlapping community detection method and two overlapping community detection methods. The main innovative results of this thesis are as follows:The first, focusing on the edges in the important significance in the detection of complex network community structure, we proposed a weight measure based on nodes intimacy and node degree(WMID). Applying this method in the community detection of a complex network, and designing non-overlapping community detection method based on Intimacy and Degree(CDID) and designing overlapping community detection method based on Intimacy and Degree(OCDID). In the community detection, Complex network will be firstly converted to a weighted network by WMID, then detecting the network. There are a large number of experiments using many groups multi-network dataset. Experimental results show that the community detection method based on WMID can quickly and efficiently detect network community structure.WMID method provides a practical, flexible and reliable method for undirected or directed, nonoverlapping or overlapping community detection.The second, because of the low complexity, fast speed, detection results close to the real society of seeds overlapping community detection method, this thesis proposed a new overlapping algorithm based on cutting and seeds expansion(OACSE). The algorithm is innovative to combined biconnected graph theory with network community detection, and by conductivity value evaluation method and the community transmission method based on neighbor nodes to expansion seed nodes and detection community. There are a large number of experiments on real network dataset. Experiments show that OACSE algorithm can effectively found community structure in small and medium-sized network. OACSE algorithms provides another feasible method for overlapping community detection research.