Explainable Predictive Process Monitoring

Abstract

The main goal of predictive process monitoring is predicting a possible outcome, execution time, and the cost of a business process by using historical data. The predictions are given at runtime, and historical data is provided in terms of an event log. Each predictive monitoring system contains predictive models which are the main part of it. Predictive models are used to make predictions and are built using information contained in the event logs. However, it is not enough just to show the prediction without giving an explanation since users want to know the rationale behind a prediction. If a person wants to take any action based on the prediction, it is definitely needed to explain the prediction in an understandable way otherwise it would be difficult to trust it. Therefore, in this thesis, we will show why explainable predictive monitoring is useful. We will do this by implementing different predictive model explanation methodologies and by investigating their application in real-life scenarios.