Automata-based Pattern Mining from Imperfect Traces
Giles Reger, Howard Barringer, David E. Rydeheard · ACM SIGSOFT Software Engineering Notes · 2015
This paper considers automata-based pattern mining techniques for extracting specifications from runtime traces and suggests a novel extension that allows these techniques to work with so-called imperfect traces i.e. traces that do not exactly satisfy the intended specification of the system that produced them. We show that by taking a so-called edit-distance between an input trace and the language of a pattern we can extract specifications from imperfect traces and identify the parts of an input trace that do not satisfy the mined specification, thus aiding the identification and location of errors in programs.