Extending Temporal Business Constraints with Uncertainty

Abstract

Monitoring and conformance checking are fundamental tasks to detect deviations between the actual and expected courses of execution in a business process. In a variety of application domains, the model capturing the expected behaviors may be intrinsically uncertain, e.g., to distinguish between standard courses of execution and exceptional but still conforming ones. Surprisingly, only very few approaches consider uncertainty as a first-class citizen in this spectrum. In this paper, we tackle this timely and challenging problem in the setting where the process of interest is described as a set of declarative, temporal business constraints. Our contribution is threefold. First, we delve into the conceptual meaning of probabilistic constraints and their semantics. Second, we argue that probabilistic constraints can be discovered from event data using existing techniques for declarative process discovery. Third, we study how to monitor probabilistic constraints, where constraints and their combinations may be in multiple monitoring states at the same time, though with different probabilities.

Declarative Process Specifications: Reasoning, Discovery, Monitoring

“…To deal with this form of uncertainty, Declare has been recently extended with probabilistic constraints [62]. In this framework, every probabilistic constraint comes with:…”

Reasoning on Labelled Petri Nets and Their Dynamics in a Stochastic Setting

“…Since multiple probabilistic constraints can be given, combined conditions on the probability of their satisfaction/violation must be obtained. To systematically handle this, we rely on the ProbDeclare language from [41], which extends the well-known Declare lan-guage [49,47] with constraint probabilities. Before defining and solving our stochastic conformance problem, we briefly recall ProbDeclare.…”

“…By exploiting a probabilistic version of LTL f as underlying temporal logic [40], the crisp semantics of Declare has been lifted to an uncertain one in [41], opening up a full range of new tasks for (probabilistic) declarative process mining [4]. In the resulting ProbDeclare framework, each constraint comes with a probability condition identifying a set of probabilities for which a trace generated by the process satisfies that constraint.…”

“…A ProbDeclare specification is a set D of probabilistic constraints over A. ◁Differently from[41,4], we do not single out crisp constraints from genuinely probabilistic constraints. This is just for simplicity of presentation, recalling that a crisp constraint φ can be represented as the corresponding probabilistic constraint ⟨φ, =, 1⟩.…”