An Algorithm Of Distributed Mining Community Structure Based On Local Information

Complex system is an important part of the real world. Complex network is a abstract ofcomplex system. Researching and mining the characters of the complex networks can helppeople to understand the complex systems. With the tendency of net society and thedevelopment of technology of compute science, people’s ability of modeling the real worldhas made substantial progress. So the researchers of network science begin to face morecomplex network model than before. Also the scientists have more and more tasks than before.Research the community structure can help us to know the complex networks clearly. Becauseof these, solving the question of detecting community structure in complex networks becomesa meaningful work.In the real world, people use more and more data in their normal life andwork, at the same time they also create more and more data which need to be handled. So howto get the meaningful data for specific populations at a reasonable time in mass data is achallengeable question. Distributed Computing is one of the main ways that people use it tohandle huge amounts of data.In our paper, firstly, we analyze the basic feature of community structure of complexnetwork then lead to the character of community structure of complex network. Secondly, weanalyze some algorithms of community structure and know the advantage and disadvantage ofthe algorithms. The algorithm of community structure based on global information need toknow the global topology of the network which has a high time and space complexity. Thecomplex network’s global topology is hard to get. For this reason our article is studying andimplementing a mining algorithm of community structure based on local information. And forimproving the performance of our algorithm, we implement it on a distributed platform. Forthe advantage of the distributed platform, our algorithm archives a balance of time and space.In our paper, we raise a new function to decide a certain node belongs to which communitywhich based on the local modality raised by Clauset. It’s used to improve the Bagrowalgorithm and we get a improved algorithm. Improved algorithm takes into account moreinformation about the nodes and gets a better result on accuracy. However, the algorithmdoesn’t increase the iterative speed. Then paper analyzes the characteristics of nodes in thelocal community and gets two heuristic methods. According to the heuristic methods we created a new algorithm. The new algorithm improves the speed of iterative while keeping theaccuracy. Then we implement the algorithm with the model of MapReduce, and run theprogram on the Hadoop platform.The result of our experiment shows us that improved two algorithms have a moreaccuracy than the raw Bagrow algorithm. The experiment on the Hadoop platform gives usthe clue that distributed computation is a useful way to solve the increasing of data quantity.