Overlapping Community Detection Algorithm Based on Spectral and Fuzzy C-Means Clustering

• Published in: Computer Supported Cooperative Work and Social Computing Vol. 917; pp. 487 - 497
• Main Authors: He, Xiaoshan, Guo, Kun, Liao, Qinwu, Yan, Qiaoling
• Format: Book Chapter
• Language: English
• Published: Singapore Springer 2019; Springer Singapore
• Series: Communications in Computer and Information Science
• Subjects: Fuzzy C-means; Overlapping community; Spectral cluster
• ISBN: 9789811330438; 9811330433
• ISSN: 1865-0929; 1865-0937
• DOI: 10.1007/978-981-13-3044-5_36

Community detection is the detection and revelation of the communities inherent in different types of complex networks, which can help people understand various functions and hidden rules of the complex networks to predict their future behavior. The spectral clustering algorithm suffers from the disadvantage of spending too much time for calculating eigenvectors, so it can’t apply in large-scale networks. This paper puts forward the overlapping community detection algorithm devised upon spectral with Fuzzy c-means clustering. Firstly, the node similarity is calculated according to the influence of attribute features on nodes. Secondly, the node similarity is combined with the Jaccard similarity to construct the similarity matrix. Thirdly, the feature decomposition is performed on the matrix by using the DPIC (Deflation-based power iteration clustering) method. Finally, the advanced version of the traditional Fuzzy c-means algorithm can find the overlapping communities. The results of experiments reveal that it can detect communities on real and artificial datasets effectively and accurately.