Pattern-based process discovery

Abstract

Service-oriented systems consist of services exchanging messages. The service interactions often only form at runtime, and implicitly define instances of a business process. Our research addresses the problem of identifying process instances, so that process analytics and optimization techniques may be applied to service-oriented systems. We introduce a novel approach to discover process instances, which takes a catalog of messaging patterns and actual message logs as input. The pattern catalog describes abstract, common service interactions, while the message logs comprise service correlation information of the concrete system under consideration. We present the required concepts, a prototype, and an experimental evaluation of our approach. Results show that pattern-based process discovery is a feasible option to discover process instances in service-oriented systems.