A Data-Driven Prediction Framework for Analyzing and Monitoring Business Process Performances

This paper presents a framework for analyzing and predicting the performances of a business process, based on historical data gathered during its past enactments. The framework hinges on an inductive-learning technique for discovering a special kind of predictive process models, which can support the run-time prediction of a given performance measure (e.g., the remaining processing time/steps) for an ongoing process instance, based on a modular representation of the process, where major performance-relevant variants of it are modeled with different regression models, and discriminated on the basis of context variables. The technique is an original combination of different data mining methods (ranging from pattern mining, to non-parametric regression and predictive clustering) and ad-hoc data transformation mechanisms, allowing for looking at the log traces at a proper level of abstraction, in a pretty automatic and transparent way. The technique has been integrated in a performance monitoring architecture, meant to provide managers and analysts (and possibly the process enactment environment) with continuously updated performance statistics, as well as with the anticipated notification of likely SLA violations. The approach has been validated on a real-life case study, with satisfactory results, in terms of both prediction accuracy and robustness.