Precise Quantitative Analysis of Probabilistic Business Process Model and Notation Workflows

Abstract

We present a framework for modeling and analysis of real-world business workflows. We present a formalized core subset of the business process modeling and notation (BPMN) and then proceed to extend this language with probabilistic nondeterministic branching and general-purpose reward annotations. We present an algorithm for the translation of such models into Markov decision processes (MDP) expressed in the syntax of the PRISM model checker. This enables precise quantitative analysis of business processes for the following properties: transient and steady-state probabilities, the timing, occurrence and ordering of events, reward-based properties, and best- and worst- case scenarios. We develop a simple example of medical workflow and demonstrate the utility of this analysis in accurate provisioning of drug stocks. Finally, we suggest a path to building upon these techniques to cover the entire BPMN language, allow for more complex annotations and ultimately to automatically synthesize workflows by composing predefined subprocesses, in order to achieve a configuration that is optimal for parameters of interest.