Research On Community Detection In Directed Neteork

Community detection has become a hotspots in the research field of complex networks, in recent years, many community detection algorithms have been proposed, but most of them have a common flaw that they were mainly suitable to the undirected networks. These algorithms are often difficult to achieve better results for directed network. Moreover, the large amount of data generated makes the traditional community detection algorithms can not meet the computational compiexity and data storage and other needs. Therefore, it is meaningful to find the community detection algorithm for large-scale directed network. The main work of this thesis are as follows:(1) The concept of complex network, the concept of community and Hadoop technology were introduced in this thesis, some representative conmmunity detection algorithms were summarized and their advantages and disadvantages were pointed out; the research status and the problems of community detection were especially studied.(2) The algorithm of directed network community detection based on similarity was studied in this thesis. The direction information of the network to guide the calculation of node simlilarity was used in this algorithm, so as to convert the topology information of the directional network into algebraic value, and then improvement of CNM algorithm by similarity,combined with the advantages of CNM algorithm to improve the accuracy of community detection.(3)Aiming at the limitation of traditional community detection algorithm in dealing with large-scale network data in a standalone, the algorithm proposed in this thesis is parallelized by using the distributed MapReduce programming model, which makes the community detection of large-scale network data realized.The similarity-based directed network community detection algorithm was used similarity to improve the CNM algorithm, which made the algorithm corresponding with the topology of networks, the calculation of similarity was guided by the direction information, making the detection of directed networks to be achieved. Experiments in standalone and distributed environments indicate that the algorithm proposed in this thesis has high accuracy and is efficient for large-scale network data processing.