Time-bounded statistical analysis of resource-constrained business processes with distributed probabilistic systems

Abstract

Business processes often incorporate stochastic decision points, either due to uncontrollable actions or because the control flow is not fully specified. Formal modeling of such business processes with resource constraints and multiple instances is hard because of the interplay among stochastic behavior, concurrency, real-time and resource contention. In this setting, statistical techniques are easier to use and more scalable than numerical methods to verify temporal properties. However, existing approaches towards simulation techniques of business processes typically rest on shaky theoretical foundations. In this paper, we propose a modular approach towards modeling stochastic resource-constrained business processes. We analyze such processes in presence of commonly used resource-allocation strategies. Our model, Distributed Probabilistic Systems (DPS), incorporates a set of probabilistic agents communicating among each other in fixed-duration real-time. Our methodology admits statistical analysis of business processes with provable error bounds. We also illustrate a number of real-life scenarios that can be modeled and verified using this approach.