An Incremental Solution for Content-Structural Graph Clustering

Researchers have always been interested in graph nodes clustering based on content or structure. But less attention has been paid to clustering based on both structure and content. But a content-structural clustering is needed in information networks like social networks. In this paper, the ICS-Cluster algorithm is proposed which takes into consideration both the structure and content aspects of the nodes. The purpose of this approach is to gain a coherent internal structure (structural aspect) and homogeneous attribute values (content aspect) in the graph. In this approach firstly the graph is converted into a content-structural graph which edges’ weight show similarity between the connected nodes. Incremental clustering is done based on edges’ weight in this process the edges with the most weight is considered as clusters then the weight of connected edge to the cluster is aggregated and they’ll be one edge, the process is repeated until the algorithm reaches the number of clusters that indicated by the user. ICS-Cluster algorithm number of cluster is indicated by the user. Comparing ICS-Cluster with other content structural algorithm based on six criteria for measuring cluster quality shows that ICS-Cluster has good performance. These criteria contain structural criteria (Modularity, Error Link, and Density), content criterion (Average Similarity), content-structural criterion (CS-Measure) and the run time.