Title

A Quickstart in Frequent Structure Mining can make a Difference

Authors

Siegfried Nijssen

Abstract

Recently a broad range of data mining algorithms has been introduced for mining frequent substructures in databases of graphs, trees and sequences. In this talk, we will introduce our new GrAph, Sequence and Tree ExtractiON (GASTON) algorithm, which combines an efficient path, tree and graph miner into one algorithm. The basis of the algorithm is that in practical databases most frequent structures are actually not complex structures, but more simple structures like free trees. An algorithm with competitive performance can be obtained by exploiting this quickstart observation. We will provide experimental results obtained by mining large molecular databases from the NCI.


Last modified: $Date: 2004/03/02 08:56:25 $ (UTC)