Repairing Activity Start Times to Improve Business Process Simulation

Abstract

Business Process Simulation (BPS) is a common technique to estimate the impact of business process changes, e.g. what would be the cycle time of a process if the number of traces increases? The starting point of BPS is a business process model annotated with simulation parameters (a BPS model). Several studies have proposed methods to automatically discover BPS models from event logs -- extracted from enterprise information systems -- via process mining techniques. These approaches model the processing time of each activity based on the start and end timestamps recorded in the event log. In practice, however, it is common that the recorded start times do not precisely reflect the actual start of the activities. For example, a resource starts working on an activity, but its start time is not recorded until she/he interacts with the system. If not corrected, these situations induce waiting times in which the resource is considered to be free, while she/he is actually working. To address this limitation, this article proposes a technique to identify the waiting time previous to each activity instance in which the resource is actually working on them, and repair their start time so that they reflect the actual processing time. The idea of the proposed technique is that, as far as simulation is concerned, an activity instance may start once it is enabled and the corresponding resource is available. Accordingly, for each activity instance, the proposed technique estimates the activity enablement and the resource availability time based on the information available in the event log, and repairs the start time to include the non-recorded processing time. An empirical evaluation involving eight real-life event logs shows that the proposed approach leads to BPS models that closely reflect the temporal dynamics of the process.