The 3rd International Workshop on Data, Text, Web, and Social Network Mining

Research Article

Detecting Overlapping Community Structures with PCA Technology and Member Index

  • @INPROCEEDINGS{10.4108/eai.18-6-2016.2264206,
        author={Peiyan Yuan and Wei Wang and Mingyang Song},
        title={Detecting Overlapping Community Structures with PCA Technology and Member Index },
        proceedings={The 3rd International Workshop on Data, Text, Web, and Social Network Mining},
        publisher={ACM},
        proceedings_a={DTWSM},
        year={2016},
        month={12},
        keywords={social networks overlapping community structure principle component analysis membership index},
        doi={10.4108/eai.18-6-2016.2264206}
    }
    
  • Peiyan Yuan
    Wei Wang
    Mingyang Song
    Year: 2016
    Detecting Overlapping Community Structures with PCA Technology and Member Index
    DTWSM
    ACM
    DOI: 10.4108/eai.18-6-2016.2264206
Peiyan Yuan1,*, Wei Wang1, Mingyang Song2
  • 1: School of Computer and Information Engineering, Henan Normal University
  • 2: School of Computer and Information Engineering Henan Normal University
*Contact email: peiyan@htu.cn

Abstract

The community structures reflect the basic property of social networks and the key point is to detect them effectively. Traditional solutions such as the nonnegative matrix factorization approach have a high time and space complexity, resulting in poor scalability. In this paper, we propose PCA-MI, a novel method to detect the overlapping community structures. Firstly, we use principle component analysis (PCA) technology to extract the key features of network information, and then employ membership index (MI) to classify nodes. Experimental results show that our approach can fast identify the overlapping community structures and achieve approximate Module-Q value as the traditional algorithms.