Using event logs to model interarrival times in business process simulation

Abstract

The construction of a business process simulation (BPS) model requires significant modeling efforts. This paper focuses on modeling the interarrival time (IAT) of entities, i.e. the time between the arrival of consecutive entities. Accurately modeling entity arrival is crucial as it influences process performance metrics such as the average waiting time. In this respect, the analysis of event logs can be useful. Given the limited process mining support for this BPS modeling task, the contribution of this paper is twofold. Firstly, an IAT input model taxonomy for process mining is introduced, describing event log use depending on process and event log characteristics. Secondly, ARPRA is introduced and operationalized for gamma distributed IATs. This novel approach to mine an IAT input model is the first to explicitly integrate the notion of queues. ARPRA is shown to significantly outperform a benchmark approach which ignores queue formation.