Process Mining for Predictive Monitoring

Abstract

Predictive process monitoring combines historical data from complete cases to predict information during the execution of running cases. We can use the outcome predictions to influence running cases. The prediction can indicate, for instance, the remaining time a patient will stay at the hospital or if he will need a specific exam. When building a predictive model, it is essential to determine the outcome prediction, the information available to input the model, and the most suitable approach. In complex processes, unstructured or context information related to the process can also be used combined with event data from event logs. Providing information where we can still act can bring interesting gains while monitoring different business processes.