Mining and re-engineering transactional workflows for reliable executions

A continuous evolution of business process parameters, constraints and needs, hardly foreseeable initially, requires from the business process management systems a continuous design and a reliable process model. In this paper, we are interested in developing a reactive design through a process log analysis ensuring process re-engineering and execution reliability. We propose to analyse workflow logs to discover workflow transactional behavior and to improve and correct related recovery mechanisms subsequently. Our approach starts by collecting workflow logs. Then, we build, by statistical analysis techniques, an intermediate representation specifying elementary dependencies between activities. These dependencies are refined to mine the transactional workflow model. The analysis of the discrepancies between the discovered model and the initially designed model enables us to detect design gaps, concerning particularly the recovery mechanisms. Thus, based on this mining step, we apply a set of rules on the initially designed workflow to improve workflow reliability.