Community Discovery Algorithms For Networks Based On Density Clustering

Community detection is an important research content of data mining for networks,which is used to explore the potential of cluster structure in the networks.It can help people to identify specific network groups,select the marketing strategies,and product recommendations,etc.It has been widely used in biological networks,social media,commercial transactions and transportation of logistics.The object of research is the network data.In the paper,we make a systematic study based on density clustering technology for global and local communities.The main contents of this thesis are summarized as follows:(1)For network data,the density algorithm is proposed which is an extension of CFSFDP.We propose the new density and dissimilarity measures for nodes so that the algorithm can be applied to the community detection.The new extended algorithm not only inherits the advantages of the original algorithm,but also can find the cluster structure of any shape.Besides,it can quickly and effectively handle network data.Finally,the experimental analysis on a large number of real network data and network data shows the effectiveness of the new algorithm.(2)As the scalability of the network data increases,the cost of global community detection is very expensive.In addition,in many cases,users tend to focus on the local community.Therefore,based on the content of the first study,we propose a local search algorithm which begins from a given node to search for the center and the border points of its community.Experiments demonstrate that this algorithm can efficiently find the local communities in the network.(3)We design and develop an experimental system for community detection.This system includes the data importation,the algorithm selection,the evaluation function selection,community detection.The system integrates the traditional algorithms and the new algorithms proposed in this paper.The tested network data includes the real and synthetic networks.The system has good availability and scalability.The above mentioned contributions has further enriched the research on the community detection,and provide a new technology support for the studies of the related fields on the network data mining and knowledge discovery.