An improved multi objective evolutionary algorithm for detecting communities in complex network with graphlet measure
In the past few years, community detection has garnered much attention due to its significant role in analysing the structures and functions of complex networks. Despite many efforts to design an effective community structure formula, the definition is still general, depends solely on the intra- and...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Published: |
Elsevier B.V.
2020
|
| Subjects: | |
| Online Access: | http://eprints.utm.my/id/eprint/92629/ http://dx.doi.org/10.1016/j.comnet.2019.107070 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Universiti Teknologi Malaysia |
| Summary: | In the past few years, community detection has garnered much attention due to its significant role in analysing the structures and functions of complex networks. Despite many efforts to design an effective community structure formula, the definition is still general, depends solely on the intra- and inter-connections of the individual nodes, and lacks complete reflection of inherent topological properties. In this paper, we continue the research line in solving community detection problem by improving the evolutionary algorithm's predictive power to address community detection challenges and transcend the limitations in earlier studies thru exploiting of strong-theoretically grounded topological properties derived from a network like graphlet measure, in terms of graphlet degree signatures and signature similarities, in method's main components. To this end, the contribution of this study is summarised in two-fold. Firstly, redefining the community detection issue as a Multi-Objective Optimisation (MOO) model, which can optimise the neighbourhood relationships along with the signature similarity measure. Secondly, proposing a new heuristic mutation operator to guide the search after considering the intra- and inter-neighbourhood topological similarity scores of the community structure, in an attempt to offer a positive collaboration with the MOO model. The systematic experiments on two benchmark networks and ten real-world networks have demonstrated the effectiveness and robustness of the proposed model in defining community structure compared with other state-of-the-art models. Furthermore, the results show that the proposed heuristic mutation operator can also improve the predictive power of the competitor MOO models in terms of convergence reliability and convergence velocity. |
|---|