Compressing Graph Data by Leveraging Domain Independent Knowledge

Abstract

Graphs are used to solve many problems in the real world.At the same time size of the graphs presents a complexscenario to analyze essential information that they contain.Graph compression is used to understand high level structureof the graph through improved visualization. In this work,we introduce CRADLE (CompRessing grAph data with Domainindependent knowLEdge), a novel method based onknowledge rule called netting, which reports the number ofexternal networks for each instance of the substructure. Byfinding such substructures with more number of external networkswe can judiciously improve the compression rate. Weempirically evaluate our approach using synthetic as well asreal-world datasets. We compare CRADLE with baseline approaches.Our proposed approach is comparable in compressionrate, search space, and runtimes to other well-knowngraph mining approaches.