Belief network discovery from event logs for business process analysis

Business processes are a main part of any organization therefore it is essential to improve their execution. Analysis of real process data can provide useful insights. Process mining techniques can be applied to event logs containing data related to business process execution to discover business processes and their behaviour therefore improving decision support. This paper presents an approach to discover probabilistic belief network from event logs, which focuses on domain-specific data contained in the logs for the analysis of business process behaviour. For evaluation purposes, the approach is applied to predict the business process execution. Experiments presented in the paper showcase practical application of the approach for synthetic and real-life logs. Obtained results prove that the approach is suitable for follow-up activity prediction and the nature of the approach allows for it to be extended for other use cases, such as anomaly detection or business process simulation.