A formal model for process context learning

Abstract

Process models are considered to be a major asset in modern business organizations. They are expected to apply to all the possible business contexts in which the process may be executed, however not all of these are known a priori. Instead of identifying all contexts before the process is established, we propose to learn from runtime experience which contextual properties should be taken into account by the process model. We propose a model and an associated procedure for identifying and learning the relevant context categories of a process out of runtime experience. We postulate that the context of a process, namely, properties of the specific business case and environmental events, affects its execution and outcomes. However, when a process is launched, the exact effect and affecting variables are not necessarily known. Our approach aims at categorizing possible environmental conditions and case properties into context categories which are meaningful for the process execution. This is achieved by a context learning framework, presented in the paper. © 2010 Springer-Verlag.