Abstract
Many community discovery algorithms add attribute information of nodes to further improve the quality of community division in the complex network with redundant and discrete data, but these algorithms lack of multi-dimensional information, such as users' interests in social networks, social relations, geography and education background, in addition to topological structure and attribute information. Therefore,this paper proposes a Multi-dimensional Information Fusion Community Discovery(MIFCD) method. Firstly, based on the idea of label propagation, link information and attribute information are combined to get link weights between nodes. Secondly, link weights are added to the topology potential to divide the sub group communities. Finally, the sub group communities are combined by using the distance information and attribute information of the core nodes between communities. In order to verify the effectiveness of the algorithm proposed in this paper, the algorithm is compared with six community partition algorithms which only consider the link information of nodes and consider the two kinds of information of node attributes and links. Experiment results on eight social networks show that this method can effectively improve the quality of community classification in both attribute communities and non-attribute communities by analyzing four evaluation indexes: improved modular degree, information entropy, community overlap degree and comprehensive index.
Highlights
Many complex systems can be regarded as complex and abstract networks composed of vertices and links or edges, such as computer networks, information networks, social networks, biological networks, etc. [1]–[4] community detection problems are of great significance to the study of complicated systems and our daily life
This paper proposes a Multi-dimensional Information Fusion Community Discovery(MIFCD) method, VOLUME 8, 2020 which is a topological potential community discovery algorithm combined with label propagation
Because of the characteristics of the actual network data, such as redundant relationship, a large number of data storage, discrete data distribution and so on, the algorithm of community division using local nodes with the highest topological potential as the core nodes of the community is easy to cause high degree of community overlap and large number of communities.after the sub-group community is divided, the community merging using the distance between the sub-group nodes and the attribute features solves the above problems while ensuring the tightness of the links between the nodes in the community and the relevance of the attributes
Summary
Many complex systems can be regarded as complex and abstract networks composed of vertices and links or edges, such as computer networks, information networks, social networks, biological networks, etc. [1]–[4] community detection problems are of great significance to the study of complicated systems and our daily life. Community detection based on labels uses randomly generated labels of each node and refreshes the labels of each node in rounds until the labels of each node no longer change, such as the NGLPA algorithm [21], ELPA-ACO algorithm [22], LPPB algorithm [23], etc These algorithms take into account the attribute information of nodes to make the community modular, but due to the characteristics of the real network data, such as redundant relationship, large amount of data storage, discrete data distribution and so on, the community is divided into a high degree of overlap and a large number of communities.how to make full use of this complex multi-dimensional information to improve the quality of results has become an important problem for community detection. The algorithm proposed in this paper divides social networks into communities based on topological potential structure. Community detection based on the domain topological potential uses node connection information to build the topological potential field in which we can partition the community.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
