A probabilistic-based approach to process model discovery

Process discovery is crucial for understanding how business operations are performed and how to improve them. The opportunity to discover process models exists given that many systems underlying the execution of process steps log their execution times. However, there are many challenges to discover the actual processes particularly complex ones and without making unrealistic assumptions. In this paper we present a novel probabilistic-based approach to discover high quality process models of any complexity. The approach has a series of steps to discover links between nodes corresponding to execution dependencies between tasks and at the end it ranks these links according to their probabilities of actually existing and classifies them according to their type. In this paper we formulate the process discovery problem, describe the challenges and describe our solution.