Automating the Discovery of AS-IS Business Process Models: Probabilistic and Algorithmic Approaches

Abstract

In the current corporate environment, business organizations have to reengineer their processes to ensure that process performance efficiencies are increased. This goal has lead to a recent surge of work on Business Process Reengineering (BPR) and Workflow Management. While a number of excellent papers have appeared on these topics, all of this work assumes that existing (AS-IS) processes are known. However, as is also widely acknowledged, coming up with AS-IS process models is a nontrivial task, that is currently practiced in a very ad-hoc fashion. With this motivation, in this paper, we postulate a number of algorithms to discover, i.e., come up with models of, AS-IS business processes. Such methods have been implemented as tools which can automatically extract AS-IS process models. To the best of our knowledge, no such work exists in the BPR and workflow domain. We back up our theoretical work with a case study that illustrates the applicability of these methods to large real-world problems. We draw on previous work on process modeling and grammar discovery. This work is a requisite first step in any reengineering endeavor. Our methods, if adopted, have the potential to severely reduce organizational costs of process redesign.