Process Model Modularization by Subprocess Discovery

Most companies exploit information systems to manage their business processes. Logs generated by such systems might be used to automatically learn models of such processes, e.g. for analysis and conformance checking purposes. Since logs are often not generated specifically for this purpose, the reported activities might be too fine-grained, leading to very complex and incomprehensible ('spaghetti') models. Modularization is a way to improve understandability and reusability of the models. This work proposes an approach to automatically discover modules, in the form of subprocesses, in unstructured processes, using the WoMan framework. Experimental results on synthetic data and models are promising.