Prescriptive Process Analytics using Counterfacts

In the age of technology integration and big data, the capability to extract value from abundant data has become paramount. Business Process Management (BPM), a field focused on the management of complex business processes, gives rise to a vast amount of event data. The subfield of “process mining” harnesses this data, acting as a conduit between BPM and data mining. Prescriptive analytics are a technique that uses data science techniques to provide actionable steps to improve a running process instance. This thesis delves into the realm of prescriptive process analytics, spotlighting the use of counterfactuals. Building upon established predictive frameworks, it seeks to develop a domain-agnostic prescriptive analysis mechanism. The primary aim is to generate recommendations that not only suggest the next-best activity, but also pinpoint the optimal resource to undertake it, all in a bid to optimize a predefined Key Performance Indicator (KPI). By comparing the efficacy of our proposed framework with existing methodologies on real-world datasets, we aim to underscore the significance and potential of our approach. The research unfolds through multiple chapters that elucidate foundational principles, state-of-the-art methods, our unique framework, and its evaluation through two distinct case studies. The hope is to chart a course for future endeavors in the domain, cementing the importance of prescriptive process analytics and its transformative impact on the business landscape.