A new method based on graph transformation for FAS mining in Multi-graph collections

Abstract

Currently, there has been an increase in the use of frequent approximate subgraph (FAS) mining for different applications like graph classification. In graph classification tasks, FAS mining algorithms over graph collections have achieved good results, specially those algorithms that allow distortions between labels, keeping the graph topology. However, there are some applications where multi-graphs are used for data representation, but FAS miners have been designed to work only with simple-graphs. Therefore, in this paper, in order to deal with multi-graph structures, we propose a method based on graph transformations for FAS mining in multi-graph collections.