A Predictive Learning Framework for Monitoring Aggregated Performance Indicators over Business Process Events

In many application contexts, a business process' executions aresubject to performance constraints expressed in an aggregated form,usually over predefined time windows, and detecting a likely violationto such a constraint in advance could help undertake correctivemeasures for preventing it. This paper illustrates a prediction-awareevent processing framework that addresses the problem of estimatingwhether the process instances of a given (unfinished) windowwwill violate an aggregate performance constraint, based on the continuouslearning and application of an ensemble of models, capableeach of making and integrating two kinds of predictions: singleinstancepredictions concerning the ongoing process instances ofw, and time-series predictions concerning the "future" process instancesofw (i.e. those that have not started yet, but will start by theend of w). Notably, the framework can continuously update the ensemble,fully exploiting the raw event data produced by the processunder monitoring, suitably lifted to an adequate level of abstraction.The framework has been validated against historical eventdata coming from real-life business processes, showing promisingresults in terms of both accuracy and efficiency.