Investigating the Influence of Data-Aware Process States on Activity Probabilities in Simulation Models: Does Accuracy Improve?

Business process simulation enables analysts to run a process in different scenarios, compare its performances and consequently provide indications on how to improve a business process. Process simulation requires one to provide a simulation model, which should accurately reflect reality to ensure the reliability of the simulation findings. An accurate simulation model passes through a correct stochastic modelling of the activity firings: activities are associated with the probability of each to fire. Literature determines these probabilities by looking at the frequency of the activity occurrences when they are enabled. This is a coarse determination, because this way does not consider the actual process state, which might influence the probabilities themselves (e.g., a thorough loan assessment is more likely for larger loan requests). The process state is as a faithful abstraction of the process instance execution so far, including the process-variable values, the activity firing history, etc. This paper aims to investigate how process states can be leveraged to improve activity firing probabilities. A technique has been put forward and compared with the baseline where basic branching probabilities are employed. Experimental results show that, indeed, business simulation models are more accurate to replicate the real process' behavior.