Using Action Networks to Detect Change in Repetitive Patterns of Action

We report on a simple method for detecting changes in repetitive patterns of action. The method involves computing the correlation of action networks before and after a focal date that is moved through an event log that represents the history of a process. Unlike process mining methods that focus on identifying an accurate model of a process at one point in time, we use a simplified model of the process to detect changes over time. The method can be used to objectively identify the date on which a process change occurred and also the relative magnitude of the change. We demonstrate the method with longitudinal data from two different organizations