Prof. Marlon Dumas University of Tartu, Estonia
Assoc. Prof. Fabrizio Maria Maggi University of Tartu, Estonia
Prof. Donato Malerba, PhD University of Bari, Italy
Prof. Dr. Myra Spiliopoulou University of Magdeburg, Germany
Recent years have witnessed a growing adoption of machine learning techniques for business improvement across various fields. Among other emerging applications, organizations are exploiting opportunities to improve the performance of their business processes by using predictive models for runtime monitoring. Such predictive process monitoring techniques take an event log (a set of completed business process execution traces) as input and use machine learning techniques to train predictive models. At runtime, these techniques predict either the next event, the remaining time, or the final outcome of an ongoing case, given its incomplete execution trace consisting of the events performed up to the present moment in the given case. In particular, a family of techniques called outcome-oriented predictive process monitoring focuses on predicting whether a case will end with a desired or an undesired outcome. The user of the system can use the predictions to decide whether or not to intervene, with the purpose of preventing an undesired outcome or mitigating its negative effects. Prescriptive process monitoring systems go beyond purely predictive ones, by not only generating predictions but also advising the user if and how to intervene in a running case in order to optimize a given utility function. This thesis addresses the question of how to train, evaluate, and use predictive models for predictive and prescriptive monitoring of business process outcomes. The thesis proposes a taxonomy and performs a comparative experimental evaluation of existing techniques in the field. Moreover, we propose a framework for incorporating textual data to predictive monitoring systems. We introduce the notion of temporal stability to evaluate these systems and propose a prescriptive process monitoring framework for advising users if and how to act upon the predictions. The results suggest that the proposed solutions complement the existing techniques and can be useful for practitioners in implementing predictive process monitoring systems in real life.