Remaining Cycle Time Prediction in Business Processes with TPOT Regression for Automated Machine Learning

Process mining allows the extraction of useful information from the event logs and historical data of business processes. Predictive business process monitoring is a sub-field of process mining and deals with runtime methods aimed at generating predictive models about various objectives related to a process instance by taking advantage of Machine Learning (ML) and deep learning algorithms. However, despite the large amount of research works dealing with predictions about next activity prediction and remaining time of running process instances, the objective of cycle time prediction has been overseen although it can provide higher level insights targeted the strategic planning. At the same time, Automated Machine Learning (AutoML) has not been investigated in the predictive business process monitoring domain in order to automate the time-consuming process of designing and optimizing ML pipelines. Therefore, motivated by its promising results in other domains, we propose an approach for remaining cycle time prediction in business processes based on AutoML, and specifically on the Tree-Based Pipeline Optimization Tool (TPOT) method. The results demonstrate that the proposed approach achieves high accuracy, while it minimizes human intervention in the configuration of ML pipelines.