Divide and Conquer: A Tool Framework for Supporting Decomposed Discovery in Process Mining

Abstract

Process mining has been around for more than a decade now, and, in that period, several discovery algorithms have been introduced that work fairly well on average-sized event logs, that is, event logs that contain ∼50 different activities. Nevertheless, these algorithms have problems dealing with big event logs, that is, event logs that contain 200 or more different activities. For this reason, a generic approach has been developed which allows such big problems to be decomposed into a series of smaller (say, average-sized or even smaller) problems. This approach offers formal guarantees for the results obtained by it and makes existing algorithms also tractable for larger logs. As a result, discovery problems may become feasible, or may become easier to handle. This paper introduces a tool framework, called Divide And Conquer that fully supports this generic approach and that has been implemented in ProM 6. Using this novel framework, this paper demonstrates that significant speed-ups can be achieved for discovery. This paper also discusses the fact that decomposition may lead to different results, but that this may even turn out to have a positive effect.