Research On Community Detection Technology In Complex Network

Data mining in complex networks has become one of the hot researches. Complex network has multiple features. Besides small world effect and scale-free effect, community structure has been considered to be one of the most important characteristics in complex networks. General speaking, community structure is the sets of nodes that are densely connected among themselves and have weaker connections to the other communities in complex networks. Community detection is conducive to understanding the real-world networks and analyzing the features of various complex systems. In this paper, we focus on the problem of community detection in complex networks. Specifically, our research mainly includes the following aspects:Firstly, it inspired by the real-world phenomena. During the formation of community structure, community structure is established from a small part of the user group which gradually developed into a community during the evolution process. The initial users can be thought as a small core in community. Based on the evolution process we propose the community detection method includes three steps: detecting core nodes, finding core community and establishing global community structure. Experimental on several synthetic networks and real networks illustrate the effectiveness of our approach.Secondly, as networks grew in size, a fast and efficient community detection method is particularly important. In this paper, we use nodes own element and propose a novel community detection algorithms based on attribute propagation through structural and attribute similarities. Experiments on real-world networks show our approach is more efficient and can detect superior communities compared to popular baselines.Finally, this paper provides an important insight to state the problem of community detection. We reveal the relationship between influence spread and community structure in essence by describing them. We propose a new cascade model IC-A based on real action logs. On the basis, we use the model to filter the seed set. Then, we tackle the problem of community detection. Experimental results show that it is not only entirely feasible to detect community in the way of influence spread but also the detected communities can promote the spread.