Towards Event Log Management for Process Mining - Vision and Research Challenges

Abstract

Organizations act in dynamic and constantly changing business environments, as the current times unfortunately illustrate. As a consequence, business processes need to be able to constantly adapt to new realities. While the dynamic nature of business processes is hardly ever challenged, the complexity of processes and the information systems (IS) supporting them make effective business process management (BPM) a challenging task. Process mining (PM) is a maturing field of data-driven process analysis techniques that addresses this challenge. PM techniques take event logs as input to extract process-related knowledge, such as automatically discovering and visualizing process models. The popularity of PM applications is growing in both industry and academia and the integration of PM with machine learning, simulation and other complementary trends, such as Digital Twins of an Organization, is gaining significant attention. However, the success of PM is directly related to the quality of the input event logs, thus the need for high-quality event logs is evident. While a decade ago the PM manifesto already stressed the importance of high-quality event logs, stating that event data should be treated as first-class citizens, event logs are often still considered as “by-products” of an IS. Even within the PM research domain, research on event logs is mostly focused on ad-hoc preparation techniques and research on event log management is critically lacking. This paper addresses this research gap by positioning event logs as first-class citizens through the lens of an event log management framework, presenting current challenges and areas for future research.