Estimation and characterization of activity duration in business processes

Process-aware information systems are typically used to log events in a variety of domains (e.g. commercial, logistics, healthcare) de- scribing the execution of business processes. The analysis of such logs can provide meaningful knowledge for organizations to improve the quality of their services as well as their efficiency. The prediction of activity du- rations, based on historic data from execution logs, allows the creation of feasible plans for business processes. However, a problem arises when there are discrepancies between execution logs and the actual execution. When event logs are partially human-generated there is an underlying uncertainty related to the time at which events (recorded by means of user interaction) are logged. If not taken into account, this uncertainty can lead to wrong predictions of activity durations. In this paper, we focus on creating assumptions to estimate activity durations and anal- yse their impact in the stochastic characterization. A partially human- generated logistics database is used as example.