Modern enterprises increasingly use the workflow paradigm to prescribe how business processes should be performed. Processes are typically modeled as annotated activity graphs. We present an approach for a system that constructs process models from logs of past, unstructured executions of the given process. The graph so produced conforms to the dependencies and past executions present in the log. By providing models that capture the previous executions of the process, this technique allows easier introduction of a workflow system and evaluation and evolution of existing process models. We also present results from applying the algorithm to synthetic data sets as well as process logs obtained from an IBM Flowmark installation.

Workflow management systems are finding wide applicability in small and large organizational settings. In this paper, we briefly review four large-scale applications to gauge their modeling and run-time requirements. Advanced transaction models (ATM) focus on maintaining data consistency and have provided solutions to many problems such as correctness, consistency, and reliability in transaction processing and database management environments. While such concepts have yet to be solved in the domain of workflow systems, database researchers have proposed to use, or attempted to use ATMs to model workflows. In this paper, we argue that workflow requirements in large-scale enterprise-wide applications involving heterogeneous and distributed environments far exceed the modeling and functionality support provided by ATMs, and suggest that an ATM is unlikely to provide a primary basis for workflow modeling and subsequent management. We have also presented various connotations of the term transaction that exist in the real-world organizational processes. Finally, we point out the need for looking beyond ATMs and using a multi-disciplinary approach for modeling large-scale workflow applications of the future.