Table Of ContentJosep Carmona
Boudewijn van Dongen
Andreas Solti
Matthias Weidlich
Conformance
Checking
Relating Processes and Models
Conformance Checking
Josep Carmona • Boudewijn van Dongen (cid:129)
Andreas Solti (cid:129) Matthias Weidlich
Conformance Checking
Relating Processes and Models
123
JosepCarmona BoudewijnvanDongen
DepartmentofComputerScience DeptofMathematicsandComputerScience
UniversitatPolite`cnicadeCatalunya TechnischeUniversiteitEindhoven
Barcelona,Spain Eindhoven,TheNetherlands
AndreasSolti MatthiasWeidlich
InstituteforInformationBusiness DepartmentofComputerScience
ViennaUniversityofEconomics Humboldt-Universita¨tzuBerlin
andBusiness Berlin,Germany
Vienna,Austria
ISBN978-3-319-99413-0 ISBN978-3-319-99414-7 (eBook)
https://doi.org/10.1007/978-3-319-99414-7
LibraryofCongressControlNumber:2018956267
©SpringerNatureSwitzerlandAG2018
Thisworkissubjecttocopyright.AllrightsarereservedbythePublisher,whetherthewholeorpartof
thematerialisconcerned,specificallytherightsoftranslation,reprinting,reuseofillustrations,recitation,
broadcasting,reproductiononmicrofilmsorinanyotherphysicalway,andtransmissionorinformation
storageandretrieval,electronicadaptation,computersoftware,orbysimilarordissimilarmethodology
nowknownorhereafterdeveloped.
Theuseofgeneraldescriptivenames,registerednames,trademarks,servicemarks,etc.inthispublication
doesnotimply,evenintheabsenceofaspecificstatement,thatsuchnamesareexemptfromtherelevant
protectivelawsandregulationsandthereforefreeforgeneraluse.
Thepublisher,theauthorsandtheeditorsaresafetoassumethattheadviceandinformationinthisbook
arebelievedtobetrueandaccurateatthedateofpublication.Neitherthepublishernortheauthorsor
theeditorsgiveawarranty,expressorimplied,withrespecttothematerialcontainedhereinorforany
errorsoromissionsthatmayhavebeenmade.Thepublisherremainsneutralwithregardtojurisdictional
claimsinpublishedmapsandinstitutionalaffiliations.
ThisSpringerimprintispublishedbytheregisteredcompanySpringerNatureSwitzerlandAG
Theregisteredcompanyaddressis:Gewerbestrasse11,6330Cham,Switzerland
Foreword
Conformance checking is an important—but also challenging—topic in process
mining. Most people who see process mining for the first time are dazzled by
theprocessdiscoverycapabilitiesoftoday’sprocessminingtools.However,when
people really start to use process mining, more detailed questions emerge, and it
is no longer sufficient to look at fancy process diagrams composed of boxes and
arrows.Initially,onlyacademictoolslikeProMsupportedconformancechecking.
However,inthelasttwoyears,alsocommercialtoolsstartedtosupportconformance
checking,andthisisagreatdevelopment.Moreover,itiswonderfulthatthereisnow
thisbookdedicatedtoconformancecheckingwrittenbymydearcolleaguesJosep
Carmona,BoudewijnvanDongen,AndreasSolti,andMatthiasWeidlich.Thebook
nicely complements my earlier Springer book “Process Mining: Data Science in
Action” by diving deep into the topic of conformance checking and giving it the
attentionitneeds.
Process mining relates event data and process models and bridges the gap
between data science and process science. It is a very natural, and seemingly
obvious, combination of forces. Most organizations have a continuous desire to
improvetheir processesand with the availability of data it makessense to do this
in an evidence-based manner. On the one hand, it aligns well with the current
excitement about Big data, machine learning, and artificial intelligence. On the
other hand, the techniques used are very different as demonstrated in this book.
We already had two “AI Winters” (1974–1980and 1987–1993),and I’m sure the
third one is not far away. Therefore, it is important to see that process mining is
verydifferentandalsoveryconcrete(asopposedtosomeofthehypedAItopics).
When people see process miningfor the first time, they wonderwhy they did not
hearaboutthisbefore.Thisbookwillhopefullycontributetoabetterpositioningof
thetopic.
ProcessminingresearchatEindhovenUniversityofTechnologystartedin1999.
The focus in the first year was on process discovery, and algorithms like the α-
miner and the heuristic miner were developed. We started to systematically work
on conformance checking in 2004 in the context of Anne Rozinat’s PhD thesis.
Beforethis,weonlycompared“footprintmatrices”andimplementedadhocfitness
v
vi Foreword
functions in the context of genetic algorithms. Anne’s seminal work resulted in
the token-based replay techniques also described in this book. Conformance was
expressed in terms of missing and remaining tokens. Several other notions were
developedinthisperiod(e.g.,variantsofwhatistodaycalledprecision)andtoken-
basedreplaybecamethestandard,notonlyforconformancechecking,butalsofor
decisionminingandperformanceanalysis.However,asalreadypointedoutinAnne
Rozinat’s PhD thesis, token-based replay has several limitations. The next major
breakthroughwas realized in the contextof Arya Adriansyah’sPhD work (2009–
2013). He developed the seminal notion of alignments which is one of the main
topicsinthisbook.Alignmentsprovidea“closestpath”throughtheprocessmodel.
It can be turned into a shortest-path problem by creating a kind of synchronous
productbetweentheeventlogandtheprocessmodel.
PartIIofthisbookusesalignmentsasa centralnotionandtakesa “deepdive”
into the world of conformance checking (including linear algebraic techniques,
heuristic search, and decompositions). The different chapters demonstrate that
conformance checking is very different from traditional model-based process
analysis(e.g.,simulationand otherbusiness processmanagementtechniques)and
data-centric analysis techniques such as machine learning and data mining. The
confrontation between event data (i.e., recorded behaviour) and process models
(hand-made or discovered automatically) triggers many interesting and highly
relevantquestions.
Part III of the book zooms in on the applications of process mining and
corresponding tools. Aligning recorded behaviour and modelled behaviour is
relevant in any domain where event data are collected, including logistics, manu-
facturing, finance, healthcare, customer relationship management, e-learning, and
e-government. Moreover, decision mining and bottleneck analysis depend on it.
For example, the profession of auditors will dramatically change because of
conformancechecking.Undoubtedly,conformancecheckingwillbeinthetoolbox
of futuregenerationsof analysts,consultants,managers,processowners,software
engineers,etc.
How about the future of conformance checking? What lies ahead? There is
broad consensus among process mining experts that conformance checking will
become more important. This providestwo main challenges: (1) How to improve
the performance of conformance checking when models and logs get bigger?
(2) How to balance between precision and deliberate vagueness? Compared to
state-of-the-art discovery techniques, conformance checking techniques tend to
be time-consuming. It is not always required to compute optimal alignments and
there is room for approximations or alternative conformance notions. Existing
conformance approaches also require formal models. However, this is in stark
contrast with the informal models generated by commercials tools. When there
is not enough evidence in the data, one should not show very strict and “binary”
process-model constructs. However, if there is enough evidence in the data, one
shouldnotuseprocess-modelconstructswithoutclearsemantics.Someofthetools
resolve this by having two types of process models: one for discovery and one
for conformance checking. Obviously, this is only a temporary solution. Process
Foreword vii
miningtoolsneedtosupporthybridprocessmodelsthatareprecisewhenpossible
and vague when needed. Such models should also take into account probabilities
of process paths. These challenges show that conformance checking continues to
providewonderfulopportunitiesforresearchers.
IwouldliketocongratulateJosep,Boudewijn,Andreas,andMatthiaswithdoing
anexcellentjobinjoiningthestate-of-the-artapproachesinasinglebook.Thebook
willbeareferenceforpeopleinterestedintheintricaterelationshipbetweenmodels
andreality.Enjoyreading!
RWTHAachenUniversity Prof.Dr.Ir.WilvanderAalst
Aachen,Germany
March2018
Preface
A model is an artefact to represent a specific concept. It maps properties of the
conceptinto some abstractrepresentation,drivenbythe purposeof the model.As
such, models represent a universal means to capture concepts in a wide range of
domains:Inarchitecture,amodelmayrepresenttheoutlineofabuildingstructure;
in engineering, a model may capture the mechanics of a pump; and in computer
science, a model of a processor may correspond to the hardware supporting a
particularsysteminfrastructure.Regardlessofthespecificconceptthatiscaptured,
amodelalwayshasapragmaticfeature.Itiscreatedforaspecificpurpose,itmay
bethedesignorconstructionoftheconceptoritsanalysisandexploration.
Inthisbook,weconsidermodelsofprocesses.Aprocessisasetofactivitiesthat
areexecutedinacoordinatedmannertoachieveacertaingoal.Itisveryhardtonot
encounteraprocessinmostofthethingswedoinourdailylife:takingthecarto
gotowork,extractingmoneyfromanATM,orderinggoodsusinganonlineshop,or
visitingthedoctor—allareexamplesofprocesses.Eachoftheminvolvesactivities,
whereastheabilitytoreachaspecificgoaldependsontheireffectivecoordination.
Processmodelsareavaluablesourceofinformation.Theymaybetheresultofa
significanteffortinformalisingacomplexprocess.Thisformalisationencompasses
the decision which properties of a process should be mapped into the model. For
processes,this decisionprimarilyrelatesto the choice ofthe activitiesto consider
andthepossiblewaysinwhichtheycanbeexecuted.However,furtherinformation
may be considered, such as the involved actors, the data used or generated, or
the timing of the activities. Modelling a real process can only be done if some
abstraction is applied. To escape from unnecessary details and retain the essence
of the process, only some process properties are mapped to a model. Hence, by
definition,modellingincurssomelossofinformation.
Yet,thefactthatnotallinformationaboutaprocesscanbecapturedinamodel
isnottheonlyreasonthataddstotheuncertaintyintherelationbetweenaprocess
modeland the processitself. A modelmay also have been formalised incorrectly,
simply became outdated with respect to the process, or describe behaviour that
deviatesfromtheactualprocess.Inthepast,thisuncertaintyintherelationbetween
aprocessandamodelthereofhasbeenacriticalproblem,whichwashardtodetect.
ix
x Preface
Themassiveavailabilityofdata, however,haschangedthissituationdramatically.
Nowadays, systems record the footprints left by executions of the process, which
enable data-driven analysis. Therefore, the recorded behaviour of a process, as
manifested in so-called event logs, is now available as an important source of
information.
Conformancecheckingrelatesmodelledandrecordedbehaviourofaprocessto
each other. It provides techniques and methods to compare and analyse observed
instancesofaprocessinthepresenceofamodel,independentoftheoriginofthat
model.Conformancecheckingthereforeaimsatansweringquestions,suchas:
(cid:129) Istheprocessbeingexecutedasitisdocumentedinamodel?
(cid:129) Isthemodelofaprocessstillup-to-date?
(cid:129) Havetherebeenviolationsofrulesintherecordedinstancesofaprocess?
(cid:129) Howmuchflexibilityisallowedintheexecutionofaprocess?
Theaimofthisbookistointroducereaderstothefieldofconformancecheckingas
awholeandoutlinethefoundationsoftherelationbetweenmodelledandrecorded
behaviour.Thebookstrivesforanoverviewoftheessentialtechniquesandmethods
in this field on the intuitive level as well as an understanding through a precise
formalisationoffundamentalnotionsofconformancechecking.
Structure ofThisBook
Thisbookisstructuredinthreeparts,eachbeingguidedbyadifferentquestionand
thereforeadifferentapproachtothefieldofconformancechecking:
Part I: What are the main ideas behind conformance checking? Readers shall
benefitfromthefirstpartofthebookasaconciseandcomprehensiveoverviewof
theessentialideastorelatemodelledandrecordedbehaviourtoeachother.Thispart
servesasareferencetoassesshowconformancecheckingeffortscouldbeapplied
in particular domains. Outlining these ideas on the intuitive level, this first part
is of interest to students, academics, and practitionersalike, who have a technical
background,butarenewtothefieldofconformancechecking.
Part II: What are the state-of-the-art methods for conformance checking? The
second part of the book aims at providing readers with detailed insights into
algorithms for conformance checking. This includes the most commonly used
formalnotionsforconformancecheckingandtheirinstantiationforspecificanalysis
questions. This part aims at enabling readers to initiate research in conformance
checking.Whileallbasicdefinitionsaregivenexplicitly,readersbenefitfromprior
knowledgeinformalismsforprocessmodelling,suchasPetrinets.
PartIII:Whatarethe applicationsthatbuildontopofconformancechecking?
The third part of the book presents applications that help to make sense of
conformancecheckingresults, therebyprovidinganecessarynextstep to increase
the value of a process model. Specifically, such applications help to interpret the
Preface xi
results of conformance checking and incorporate them through enhancement and
repairtechniques.Thisincludesremarksonconformancecheckingsoftware.
TeachingUsingThisBook
In the light of the above structure, three itineraries are suggested when using the
bookasareferenceforteaching:
(cid:129) Chapters1–4and12:Thesechaptersgiveageneraldescriptionofconformance
checking,includingremarksontoolsupport,fromapracticalperspective.
(cid:129) Chapters5–9:Thispartgivesanalgorithmicviewonhowtorelatemodelledand
recordedbehaviour,focusingonformalfoundationsofconformancechecking.
(cid:129) Chapters1–6and10–12:Anend-to-endviewonconformancecheckingandits
applications,coveringformalfoundationsonthebasiclevel.
Thechaptersofthebookincorporatematerialinthefollowingform:
(cid:129) Each chapter contains teaching material in the form of exercises to allow for
effectivelearningofthetheoreticalconceptsintroduced.
(cid:129) For those chapters amenable for tool practising, a section namedIn the Lab is
included,whichintroducestoolsand/ordatasetsforexploringthecontentsofthe
chapteronrealdata.
Websiteofthebook
Check out the website for teaching
materials for the book. Throughout
thebook,youfindlinkstosupportlab
sessionsforspecificchapters.
http://www.conformancechecking.com/CC_book
Barcelona,Spain JosepCarmona
Eindhoven,TheNetherlands BoudewijnvanDongen
Vienna,Austria AndreasSolti
Berlin,Germany MatthiasWeidlich
