Resumen:
Abstract. Process-aware information systems (PAISs) are increasingly
used to provide
flexible support for business processes. The support given
through a PAIS is greatly enhanced when it is able to provide accurate
time predictions which is typically a very challenging task. Predictions
should be (1) multi-dimensional and (2) not based on a single process in-
stance. Furthermore, the prediction system should be able to (3) adapt to
changing circumstances, and (4) deal with multi-perspective declarative
languages (e.g., models which consider time, resource, data and control
flow perspectives). In this work, a novel approach for generating time
predictions considering the aforementioned characteristics is proposed.
For this, rst, a multi-perspective constraint-based language is used to
model the scenario. Thereafter, an optimized enactment plan (represent-
ing a potential execution alternative) is generated from such a model
considering the current execution state of the process instances. Finally,
predictions are performed by evaluating a desired function over this en-
actment plan. To evaluate the applicability of our approach in practical
settings we apply it to a real process scenario. Despite the high com-
plexity of the considered problems, results indicate that our approach
produces a satisfactory number of good predictions in a reasonable time.
This research has been supported by the Spanish MINECO R&D Projects Pololas TIN2016-76956-C3-2-R and PLAN MINER TIN2015-71618-R.