DESSIN: Mining Dense Subgraph Patterns in a Single Graph

Currently, a large amount of data can be best represented as graphs, e.g., social networks, protein interaction networks, etc. The analysis of these networks is an urgent research problem with great practical applications. In this paper, we study the particular problem of finding frequently occurring dense subgraph patterns in a large connected graph. Due to the ambiguous nature of occurrences of a pattern in a graph, we devise a novel frequent pattern model for a single graph. For this model, the widely used Apriori property no longer holds. However, we are able to identify several important properties, i.e., small diameter, reachability, and fast calculation of automorphism. These properties enable us to employ an index-based method to locate all occurrences of a pattern in a graph and a depth-first search method to find all patterns. Concluding this work, a large number of real and synthetic data sets are used to show the effectiveness and efficiency of the DESSIN method.

• Type: Journal
• Year: 2010
• Where: SSDBM
• Authors: Shirong Li, Shijie Zhang, Jiong Yang
• Keywords: Database, Dense Subgraph Patterns, Large Connected Graph, Protein Interaction Networks, SSDBM 2010