Predictive Monitoring of Multi-level Processes

The ever increasing use of Information Systems causes ever more information to be stored. As organizations and businesses become more efficient due to competition they need to gain competitive advantage over others. More and more companies and institutions have turned to Information Technology to find business value in a data-driven world. Modern Information Systems maintain records of process events, which correspond to real-life activities. As processes evolve and become more complex, so does the information that reflects them. In this thesis, we propose an approach to predictive monitoring of complex multi-level processes. In this context, a multi-level process consists of a highlevel parent process which spawns multiple low-level subprocesses, which have their own life cycle and run independently of one another. The author proposes constructs called milestones, which include both parentand subprocesses and are used for the predictive monitoring classification task. This approach has been validated on a real-life event log of the business-to-business change management process in place at Baltic’s largest telecommunications company Telia Estonia.