Generating Event Logs with Workload-Dependent Speeds from Simulation Models

Abstract

Abstract. Both simulation and process mining can be used to analyze operational business processes. Simulation is model-driven and very useful because different scenarios can be explored by changing the model's parameters. Process mining is driven by event data. This allows detailed analysis of the observed behavior showing actual bottlenecks, deviations, and other performance-related problems. Both techniques tend to focus on the control-flow and do not analyze resource behavior in a detailed manner. In this paper, we focus on workload-dependent processing speeds because of the well-known phenomenon that people perform best at a certain stress level. For example, the "Yerkes-Dodson Law of Arousal" states that people will take more time to execute an activity if there is little work to do. This paper shows how workload-dependent processing speeds can be incorporated in a simulation model and learned from event logs. We also show how event logs with workload-dependent behavior can be generated through simulation. Experiments show that it is crucial to incorporate such phenomena. Moreover, we advocate an amalgamation of simulation and process mining techniques to better understand, model, and improve real-life business processes.

The use of process mining in a business process simulation context: Overview and challenges

“…This relationship is explicited in the Yerkes-Dodson law of arousal, proposing an inverted U-shape relationship between arousal level and employee performance [9]. To implement the Yerkes-Dodson law, Nakatumba and van der Aalst [9] and Nakatumba et al [10] estimate a linear regression based on event log information to determine the relationship between workload and processing speed for each individual employee. As recognized by Nakatumba and van der Aalst [9], improvements to the modeling method are possible.…”

An operational support approach for Mining Unstructured Business Processes

“…In this context, several approaches have been developed. Since Process Mining Manifesto released at the end of 2011, we focused on research papers published from the beginning of 2012 year: The [9] developed a concrete implementation of operational support meta-model, based on the workflow system Declare and the ProM framework. This meta-model treats four types of queries: simple queries, compare queries, predict queries, and recommend queries.…”

Matching heterogeneous event data

“…Obviously, the events in two logs with the same name correspond to each other. All these models and logs are generated by an open source toolkit BeehiveZ 4 using existing generating approaches [18,15].…”