A One-Dimensional Kalman Filter for Real-Time Progress Prediction in Object Lifecycle Processes

Real-time monitoring of business processes offers promising perspectives to discover problems and optimisation potentials. Early detection is a key part in this endeavour. One crucial aspect of real-time monitoring is to determine the current progress of a running business process. This is particularly challenging for business processes that consist of a multitude of loosely coupled, smaller processes that interact with each other, like object lifecycle processes in data-centric approaches to business process management. In this paper, an approach to predict the remaining portion of the process path to be still executed in relation to the overall process is proposed. This prediction is based on a one-dimensional Kalman Filter. As a major benefit of this approach, real-time progress determination can start directly with the first run of the process, i.e., without need for comprehensive event log data. This becomes possible due to the procedure applied by the Kalman Filter, which requires no log data. A quantitative study with 250 progress estimations for large object lifecycle processes results in a deviation of the average estimated progress from the real progress, calculated after the completion of the process, of about 5%. This emphasises that reasonable progress predictions are possible even in the absence of an event log, as it is the case when deploying new or changed processes to the run-time system