Community Detection Using Graphical Relationships

Community detection in social networks requires a careful effort in capturing relationships between individual nodes and identifying the different communities in the network. Contemporary research in this field has so far not considered nodes to be belonging to multiple communities. In this paper, the Community Detection Problem is solved by converting the social network graph to a low-dimensional space of features is attempted. By adopting flexibility in exploring a node’s neighbourhood, overlapping communities are captured. Two kinds of node neighbourhoods based on hop distance and functionality is defined and incorporated both local and global graph representation. Our algorithm is validated by applying it to large social network datasets and visualizing the number of clusters obtained. The clusters depict the various parameters on which the communities have been separated into. The larger the number of clusters, the larger is the complexity of the communities that the dataset has been separated into. On the whole, our work presents a novel approach towards community detection by using overlapping communities in a social network graph.