Hybrid fragment mining with MoFa and FSG

In the past few years a number of different subgraph mining algorithms have been proposed. They are often used for ending frequent fragments in molecular databases. All these algorithms behave quite well when used on small datasets of not more than a few thousand molecules. However they all fail on larger amounts of data because they are either time consuming or have enormous memory requirements. We present a hybrid mining technique that overcomes the individual problems of the underlying algorithms and outperforms the individual methods impressively on large databases.