The Effect of Duration Heteroscedasticity to the Bottleneck in Business Process Discovered by Inductive Miner Algorithm

One way to do business process modelling is to use the process mining. Process mining links the gap between traditional model-based process analysis such as business process management simulation and data-centric analysis techniques such as machine learning and data mining. In process modelling, bottleneck conditions are often found. Bottlenecks conditions can be found in the process models generated using Process Mining applications such as ProM and Disco based on event log data. There is another alternative to find the bottleneck condition of the event log using a statistical approach. The alternative is to view the event log as an asset that can be explored without using a normative process model. This paper proposes a statistical test of heteroscedasticity in event log data. Then the heteroscedasticity test results from the event log are compared with the results of normative process modelling with the Inductive Miner algorithm using the Process Mining application. The comparison results show that the detected event log data having heteroscedasticity problems will ensure a bottleneck condition in the process model. The approach taken can be an alternative in evaluating the process model based on its event log.