Business process event log transformation into Bayesian belief network

Abstract

Business process (BP) mining has been recognized in business intelligence and reverse engineering fields because of the capabilities it has to discover knowledge about the implementation and execution of BP for analysis and improvement. Existing business knowledge extraction solutions in process mining context requires repeating analysis of event logs for each business knowledge extraction task. The probabilistic modelling could allow improved performance of BP analysis. Bayesian belief networks are a probabilistic modelling tool and the paper presents their application in BP mining. The paper shows that existing process mining algorithms are not suited for this, since they allow for loops in the extracted BP model that do not really exist in the event log, and presents a custom solution for directed acyclic graph extraction. The paper presents results of a synthetic log transformation into Bayesian belief network showing possible application in business intelligence extraction and improved decision support capabilities.