AN EARLY WARNING STRATEGY FOR REAL-TIME BUSINESS PROCESS MONITORING

This paper proposes a novel approach to real-time business process monitoring incorporating an early warning strategy based on expected gain analysis. First, at each monitoring instant in midcourse, an extended decision-tree algorithm estimates a probability that the ongoing process will accomplish the targeted performance after completion. Second, based on that probability, the expected gains are analyzed for two cases: stopping the running process and executing continuously until completion. If stopping is more profitable, an early warning is generated to notify probable loss after completion. Otherwise, the process is executed until the next instant. This procedure is iterated periodically over entire instants. The proposed method estimates predictors reflecting the real-time progress of a running process, especially current capabilities of accomplishing business goals. Furthermore, by means of the early warning, the monitoring system can proactively prevent probable losses rather than just react to them.