The ParMol Package for Frequent Subgraph Mining

Abstract

Mining for frequent subgraphs in a graph database has become a popular topic in the last years. Algorithms to solve this problem are used in chemoinformatics to find common molecular fragments in a database of molecules represented as two-dimensional graphs. However, the search process in arbitrary graph structures includes costly graph and subgraph isomorphism tests. In our ParMol package we have implemented four of the most popular frequent subgraph miners using a common infrastructure: MoFa, gSpan, FFSM, and Gaston. Besides the pure re-implementation, we have added additional functionality to some of the algorithms like parallel search, mining directed graphs, and mining in one big graph instead of a graph database. Also a 2D-visualizer for molecules has been integrated.