Context-Aware Change Pattern Detection in Event Attributes of Recurring Activities

Process mining bridges the gap between process management and data science by utilizing process execution data to discover and analyse business processes. This data is represented in event logs, where each event contains attributes describing the process instance, the time the event has occurred, and much more. In addition to these generic event attributes, events contain domain-specific event attributes, such as a measurement of blood pressure in a healthcare environment. Taking a close look at those attributes, it turns out that the respective values change during a typical process quite frequently, hence we refer to them as dynamic event attributes. This paper studies change patterns of dynamic event attributes by recurring process activities in a given context. We have applied the technique on two real-world datasets, MIMIC-IV and Sepsis, representing hospital treatment processes, and show that the approach can provide novel insights. The approach is implemented in Python, based on the PM4Py framework.