خوشه بندی کلونی مورچه ها با درک تناسب و انتشار فرومونی برای تشخیص جامعه در شبکه های پیچیده
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|7881||2013||13 صفحه PDF||سفارش دهید||7580 کلمه|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Physica A: Statistical Mechanics and its Applications, Volume 392, Issue 15, 1 August 2013, Pages 3260–3272
Community structure detection in complex networks has been intensively investigated in recent years. In this paper, we propose an adaptive approach based on ant colony clustering to discover communities in a complex network. The focus of the method is the clustering process of an ant colony in a virtual grid, where each ant represents a node in the complex network. During the ant colony search, the method uses a new fitness function to percept local environment and employs a pheromone diffusion model as a global information feedback mechanism to realize information exchange among ants. A significant advantage of our method is that the locations in the grid environment and the connections of the complex network structure are simultaneously taken into account in ants moving. Experimental results on computer-generated and real-world networks show the capability of our method to successfully detect community structures.
Nature, society, and many technologies can be modeled by numerous networks that involve many important and complicated interactions among individuals , such as Biological networks, Social networks, Collaboration networks, and Web networks. These networks are often called complex networks. These complex networks are found to divide naturally into communities, which are groups of nodes such that the nodes within a group are much more connected to each other than to the rest of the network. In recent years, community structure discovery has been one of the most popular research areas along with its applicability to a wide range of disciplines , ,  and . For example, cell biologists use the community structure in protein interaction networks to make sense of signal transduction cascades and metabolism, to research the inherent relationships between cellular functions and biochemical events in this area. Computer scientists are mapping the Internet and the WWW into different communities to discover topic information from Web pages and potential relationships in hyperlinks. Epidemiologists follow transmission networks through which viruses spread and try to find how to stop the spread of the virus by analyzing the transmission network structure, and so on. That is, communities are interesting and fundamental because they often correspond to functional units and reflect the in-homogeneity or locality of the topological relationships between the nodes of target systems. Thus, community structure mining is important for analyzing topological structures, comprehending network functions, recognizing hidden patterns, and forecasting the future behaviors of complex systems. With the development and popularity of complex networks, various network clustering algorithms to mine community structures have been proposed , , , , , , , , , , , , , ,  and . In Ref. , Fortunato gave a comprehensive and in-depth summary for the methods and algorithms of community detection in graphs from different aspects. Although special strategies adopted are different, most of the algorithms are mainly divided into two basic categories from the view of the search mechanism: the global optimization approach and local search approach . The first approach poses clustering as a global optimization problem, which employs an objective function to evaluate the network modularity quality and tries to find an optimal clustering result in the whole solution space , , , , ,  and . In contrast, there are no global optimization objectives in the second approach, and they carry out clustering based on some local heuristic rules , , , ,  and , such as edge betweenness, clique percolation, random walk, etc. Moreover, there are some algorithms that use a combination of these two basic methods ,  and , which can get better clustering results and higher time efficiency than each basic approach. For all that, however, how to further improve the detection accuracy, especially how to discover reasonable community structure without prior knowledge, is still a challenging problem. In this paper, we propose an Ant Colony Clustering algorithm based on Fitness perception and Pheromone diffusion for community detection in complex networks (called as ACC-FP), which is also a combination method. ACC-FP first uses a new fitness function to percept local environment and move to more comfortable locations in the grid. Then, in terms of the ants clustering quality at each iteration, it employs a pheromone diffusion model, which can faithfully simulate the volatilization character of pheromone, to realize information exchange among ants and guide ant colony to perform global searches. Experimental results and relevant comparative analyses show that the combination of local perception and global information feedback is effective to achieve high-quality clustering results. The rest of this paper is organized as follows. Section 2 introduces related research on community detection. Section 3 presents the motivation and the details of the proposed algorithm. Section 4 presents and analyzes the experimental results. Section 5 concludes this paper.
نتیجه گیری انگلیسی
The identification of community structures in complex networks is of great interest because they often reveal unknown relationships between nodes and provide useful information for unknown nodes. However, how to accurately predict community structures by computational methods is still a highly challenging issue. The paper presents an ant colony clustering algorithm with high accuracy for identifying a community in complex networks. The algorithm focuses on the strategy of ant perception and movements and the method of pheromone diffusion and updating, and searches for an optimal partitioning of the network by ant colony movements. The underlying community structure of the complex network will become clearly visible at the end of the algorithm by selectively exploring the search space, without the need to know in advance the exact number of groups. The performance of our algorithm was tested on three real-world networks, as well as on a set of computer-generated networks. Experimental results confirm the validity and advantage of this approach. Future research will aim at combining our method with other methods to improve the quality of results.