Efficient discovery of understandable declarative process models from event logs

F.M. Maggi, R.P. Jagadeesh Chandra Bose, W.M.P. Aalst, van der

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

104 Citations (Scopus)
2 Downloads (Pure)

Abstract

Process mining techniques often reveal that real-life processes are more variable than anticipated. Although declarative process models are more suitable for less structured processes, most discovery techniques generate conventional procedural models. In this paper, we focus on discovering Declare models based on event logs. A Declare model is composed of temporal constraints. Despite the suitability of declarative process models for less structured processes, their discovery is far from trivial. Even for smaller processes there are many potential constraints. Moreover, there may be many constraints that are trivially true and that do not characterize the process well. Naively checking all possible constraints is computationally intractable and may lead to models with an excessive number of constraints. Therefore, we have developed an Apriori algorithm to reduce the search space. Moreover, we use new metrics to prune the model. As a result, we can quickly generate understandable Declare models for real-life event logs.
Original languageEnglish
Title of host publicationAdvanced Information Systems Engineering (24th International Conference, CAiSE 2012, Gdansk, Poland, London, UK, June 25-29, 2012. Proceedings)
EditorsJ. Ralyté, X. Franch, S. Brinkkemper, S. Wrycza
Place of PublicationBerlin
PublisherSpringer
Pages270-285
ISBN (Print)978-3-642-31094-2
DOIs
Publication statusPublished - 2012

Publication series

NameLecture Notes in Computer Science
Volume7328
ISSN (Print)0302-9743

Cite this