AITIA-PM: Discovering the true causes of events in a process mining context

Abstract

Process mining is a research area that enables businesses to analyze and improve their processes by deriving knowledge from event logs. While pinpointing the causes of, for instance, a negative case outcome can provide valuable insights for business users, only a limited amount of research has been done to uncover causal relations within the process mining field while actively distinguishing between correlation and causality. The AITIA-PM algorithm is one of these research projects. This article updates the AITIA-PM method, which uses causality theory to measure cause-and-effect relationships in event logs. The system uses probabilistic temporal logic (PTL) to formulate hypotheses explicitly and then automatically checks them for causality using available data. More precisely, AITIA-PM is designed for process mining since it operates directly on event logs, giving users access to the information stored there, and increasing the scope for meaningful causal analysis in a process mining setting. With this addition, PTL is emphasized more as a crucial algorithmic component, and the method to control for false discovery rates (FDR) is adjusted for increased practical use. The case study shows that after the domain expert provides the search space of hypotheses, the AITIA-PM algorithm can extract valuable cause–effect insights from an event log. The search space can be flexibly defined, making AITIA-PM a powerful tool for business users. An evaluation on artificial data proves AITIA-PM is capable of extracting the causal relationships, while a demonstration on the Road Traffic Fines Management dataset shows the applicability of the algorithm on real data.