Äriprotsesside ennestusseire kasutades multiperspektiivseid a priori teadmiseid
Files
Date
2018
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Äriprotsesside ennestusseire on valdkond, mis on pühendunud käimasolevate äriprotsesside tuleviku ennustamisele kasutades selleks minevikus sooritatud äriprotsesside kohta käivaid andmeid. Valdav osa uurimustööst selles valdkonnas keskendub ainult seda tüüpi andmetele, jättes tähelepanuta täiendavad teadmised (a priori teadmised) protsessi teostumise kohta tulevikus. Hiljuti pakuti välja lähenemine, mis võimaldab a priori teadmisi kasutada LTL-reeglite näol. Kuid tõsiasjana on antud tehnika limiteeritud äriprotsessi kontroll-voole, jättes välja võimaluse väljendada a priori teadmisi, mis puudutavad lisaks kontrollvoole ka informatsiooni protsessis leiduvate atribuutide kohta (multiperspektiivsed a priori teadmised). Me pakume välja lahenduse, mis võimaldab seda tüüpi teadmiste kasutuse, tehes multiperspektiivseid ennustusi käimasoleva äriprotsessi kohta. Tulemused, milleni jõuti rakendades väljapakutud tehnikat 20-le tehisärilogile ning ühele elulisele ärilogile, näitavad, et meie lähenemine suudab pakkuda konkurentsivõimelisi ennustusi.
Predictive business process monitoring is an area dedicated to exploiting past process execution data in order to predict the future unfolding of a currently executed business process instance. Most of the research done in this domain focuses on exploiting the past process execution data only, leaving neglected additional a priori knowledge that might become available at runtime. Recently, an approach was proposed, which allows to leverage a priori knowledge on the control flow in the form of LTL-rules. However, cases exist in which more granular a priori knowledge becomes available about perspectives that go be-yond the pure control flow like data, time and resources (multiperspective a priori knowledge). In this thesis, we propose a technique that enables to leverage multi-perspective a priori knowledge when making predictions of complex sequences, i.e., sequences of events with a subset of the data attributes attached to them. The results, obtained by applying the proposed technique to 20 synthetic logs and 1 real life log, show that the proposed technique is able to overcome state-of-the-art approaches by successfully leveraging multiperspective a priori knowledge.
Predictive business process monitoring is an area dedicated to exploiting past process execution data in order to predict the future unfolding of a currently executed business process instance. Most of the research done in this domain focuses on exploiting the past process execution data only, leaving neglected additional a priori knowledge that might become available at runtime. Recently, an approach was proposed, which allows to leverage a priori knowledge on the control flow in the form of LTL-rules. However, cases exist in which more granular a priori knowledge becomes available about perspectives that go be-yond the pure control flow like data, time and resources (multiperspective a priori knowledge). In this thesis, we propose a technique that enables to leverage multi-perspective a priori knowledge when making predictions of complex sequences, i.e., sequences of events with a subset of the data attributes attached to them. The results, obtained by applying the proposed technique to 20 synthetic logs and 1 real life log, show that the proposed technique is able to overcome state-of-the-art approaches by successfully leveraging multiperspective a priori knowledge.