Clustering Improves the Exploration of Graph Mining Results

Abstract

Abstract. Mining frequent subgraphs is an area of research where we have a given set of graphs, and where we search for (connected) subgraphs contained in many of these graphs. Each graph can be seen as a transaction, or as a molecule -as the techniques applied in this paper are used in (bio)chemical analysis. In this work we will discuss an application that enables the user to further explore the results from a frequent subgraph mining algorithm. Such an algorithm gives the frequent subgraphs, also referred to as fragments, in the graphs in the dataset. Next to frequent subgraphs the algorithm also provides a lattice that models sub-and supergraph relations among the fragments, which can be explored with our application. The lattice can also be used to group fragments by means of clustering algorithms, and the user can easily browse from group to group. The application can also display only a selection of groups that occur in almost the same set of molecules, or on the contrary in different molecules. This allows one to see which patterns cover different or similar parts of the dataset.

Visualization and Grouping of Graph Patterns in Molecular Databases

“…In [4] the current setup is used to cluster data; that paper discusses an application that enables the user to further explore the results from a frequent subgraph mining algorithm, by browsing the lattice of frequent graphs. Figure 2: An example of co-occurring subgraphs from [9] with an example molecule.…”