Mining Resource Availability for Data-driven Business Process Simulation

Abstract

Business process simulation (BPS) is a set of techniques to analyze a process model regarding identified performance metrics. BPS helps analysts decide whether to apply the process model to real-life production based on a set of statistics produced by a simulator. The accuracy of the output, accordingly, the simulation process’s value withdrawn, is affected by the correctness of the business process model and the simulation parameters used as input. Therefore, data-driven simulation techniques are introduced to generate a process model from the process execution data recorded by the organization’s information system to resemble reality. However, simulation models tend to be oversimplified due to some limitations of the existing simulation tools. One of the common limitations that need to be tackled is addressed towards resource availability and behavior. For example, it is assumed that resources are always available while, in reality, they have a work schedule, could have a part-time contract or are out of reach under certain conditions. In this respect, the existing business process simulators accept timetables to specify resource availability. On the one hand, providing a resource timetable to the simulator will increase the results’ accuracy. On the other hand, a formal employee timetable most likely does not reflect the actual resource schedules. To this end, this research study presents a data-driven methodology that uses the execution log of the process under consideration to capture the reality of resource availability. The calendar discovery algorithm presented is integrated with the data-driven business process simulation tool, Simod [4]. As expected, the results show an increase in the precision of the discovered business process simulation model. The evaluation was carried out on four real-life logs as well as a synthetic data set.