## Abstract

In the previous methods of community detection, unipartite networks and bipartite networks are dealt with separately, so the type of network should be known in advance. This paper presents a vertices similarity probability (VSP) model to find community structure without the priori knowledge of the type of complex network structure. As vertices in the same community have similar properties, the VSP model uses vertices similarity to find community structure which is a unified algorithm and can be used in any network without knowing the type of network structure. As "Common neighbor index" has been proved to be an effective index for vertices similarity, it is used to measure the vertices similarity probability. Then, we give the method to determine the number of communities using matrix perturbation theory. We apply the model to find community structure in real-world networks and artificial networks. The experimental results show that the VSP model is applicable to both unipartite networks and bipartite networks, and is able to find the community structure successfully without using the type of network structure.

Original language | English |
---|---|

Pages (from-to) | 36-43 |

Number of pages | 8 |

Journal | Neurocomputing |

Volume | 130 |

DOIs | |

Publication status | Published - 23 Apr 2014 |

## Keywords

- Common neighbor index
- Community detection
- Type of network structure
- Vertices similarity probability