Data-driven and automated prediction of service level agreement violations in service compositions

TLDR: This paper discusses data-driven, statistical approaches for both, instance-level prediction (SLA compliance prediction for an ongoing business process instance) and forecasting (compliance prediction for future instances) and presents an integrated framework and numerically evaluates the approach based on a case study from the manufacturing domain.

Abstract: Service Level Agreements (SLAs), i.e., contractually binding agreements between service providers and clients, are gaining momentum as the main discriminating factor between service implementations. For providers, SLA compliance is of utmost importance, as violations typically lead to penalty payments or reduced customer satisfaction. In this paper, we discuss approaches to predict violations a priori. This allows operators to take timely remedial actions, and prevent SLA violations before they have occurred. We discuss data-driven, statistical approaches for both, instance-level prediction (SLA compliance prediction for an ongoing business process instance) and forecasting (compliance prediction for future instances). We present an integrated framework, and numerically evaluate our approach based on a case study from the manufacturing domain.