GraphClust: A Method for Clustering Database of Graphs

Abstract

Any application that represents data as sets of graphs may benefit from the discovery of relationships among those graphs. To do this in an unsupervised fashion requires the ability to find graphs that are similar to one another. That is the purpose of GraphClust. The GraphClust algorithm proceeds in three phases, often building on other tools:(1) it finds highly connected substructures in each graph;(2) it uses those substructures to represent each graph as a feature vector; and (3) it clusters these feature vectors using a standard distance measure. We validate the cluster quality by using the Silhouette method. In addition to clustering graphs, GraphClust uses SVD decomposition to find frequently co-occurring connected substructures. The main novelty of GraphClust compared to previous methods is that it is application-independent and scalable to many large graphs.

Finding patterns in labeled graphs using spectrum feature vectors in a SOM network

“…The first experiment was visually evaluated and the second was evaluated using the metric CQ defined by (10). Also the results of the second experiment were compared with GraphClust method [12] (described in section I) using the option for sparse graphs.…”

“…A SOM with a map of 5x10 neurons in its output layer was trained during 10,000 epochs, using the 381 feature vectors built by our algorithm. This experiment was also executed using algorithm GraphClust 2 [12] for sparse graph, therefore Subdue was executed in its first phase (see section 1). GraphClust was asked to generate 50 clusters.…”

“…Luo [11] proposed a process for clustering nonlabeled graphs based on "Locally Preserving Projection" (LPP) and "Principal Component Analysis" (PCA). Reforgiato et al [12] proposed a method for finding graphs that are similar to each other, called GraphClust. This algorithm executes three phases: first, it finds interesting structures using the Subdue substructure discovery system [13] for sparse graphs or algorithms GraphGrep and AllPairShortestPath [14] for dense graphs.…”

“…Size of each group i generated by: a human expert , our method based on SOM , and GraphClust method[12] .…”

Intrusion Alerts Analysis Using Attack Graphs and Clustering

“…Temporal Attack Graphs prepared from attack sequences. The graphs are clustered with the "GraphClust" [19]. Shasha et al have done research to suggest a way to perform clustering on the graphs.…”