Semantical Vacuity Detection in Declarative Process Mining

Fabrizio Maria Maggi, Marco Montali, Claudio Di Ciccio, Jan Mendling

Publikation: Beitrag in Buch/KonferenzbandBeitrag in Konferenzband


A large share of the literature on process mining based on declarative process modeling languages, like DECLARE, relies on the notion of constraint activation to distinguish between the case in which a process execution recorded in event data “vacuously” satisfies a constraint, or satisfies the constraint in an “interesting way”. This fine-grained indicator is then used to decide whether a candidate constraint supported by the analyzed event log is indeed relevant or not. Unfortunately, this notion of relevance has never been formally defined, and all the proposals existing in the literature use ad-hoc definitions that are only applicable to a pre-defined set of constraint patterns. This makes existing declarative process mining technique inapplicable when the target constraint language is extensible and may contain formulae that go beyond pre-defined patterns. In this paper, we tackle this hot, open challenge and show how the notion of constraint activation and vacuous satisfaction can be captured semantically, in the case of constraints expressed in arbitrary temporal logics over finite traces. We then extend the standard automata-based approach so as to incorporate relevance-related information. We finally report on an implementation and experimentation of the approach that confirms the advantages and feasibility of our solution.
Titel des SammelwerksBusiness Process Management - 14th International Conference, BPM 2016, Rio de Janeiro, Brazil, September 18-22, 2016. Proceedings
Herausgeber*innen Marcello La Rosa, Peter Loos, Oscar Pastor
ErscheinungsortRio de Janeiro, Brazil
VerlagSpringer Lecture Notes in Computer Science (LNCS)
Seiten158 - 175
PublikationsstatusVeröffentlicht - 2016

Österreichische Systematik der Wissenschaftszweige (ÖFOS)

  • 502050 Wirtschaftsinformatik
  • 102022 Softwareentwicklung
  • 102
  • 102001 Artificial Intelligence