Fundamentals of Business Intelligence

361 Pages·2015·7.4 MB·English

by  Wilfried Grossmann

Data-Centric Systems and Applications
Wilfried Grossmann
Stefanie Rinderle-Ma
Fundamentals of
Business
Intelligence
Data-Centric Systems and Applications
SeriesEditors
M.J.Carey
S.Ceri
EditorialBoard
A.Ailamaki
S.Babu
P.Bernstein
J.C.Freytag
A.Halevy
J.Han
D.Kossmann
I.Manolescu
G.Weikum
K.-Y.Whang
J.X.Yu
Moreinformationaboutthisseriesat
http://www.springer.com/series/5258
Wilfried Grossmann • Stefanie Rinderle-Ma
Fundamentals of
Business Intelligence
123
WilfriedGrossmann StefanieRinderle-Ma
UniversityofVienna UniversityofVienna
Vienna Vienna
Austria Austria
ISSN2197-9723 ISSN2197-974X (electronic)
Data-CentricSystemsandApplications
ISBN978-3-662-46530-1 ISBN978-3-662-46531-8 (eBook)
DOI10.1007/978-3-662-46531-8
LibraryofCongressControlNumber:2015938180
SpringerHeidelbergNewYorkDordrechtLondon
©Springer-VerlagBerlinHeidelberg2015
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.
Printedonacid-freepaper
Springer-Verlag GmbH Berlin Heidelberg is part of Springer Science+Business Media
(www.springer.com)
Foreword
IntelligentbusinessesneedBusinessIntelligence(BI).Theyneeditforrecognizing,
analyzing,modeling,structuring,andoptimizingbusinessprocesses.Theyneedit,
moreover, for making sense of massive amounts of unstructured data in order to
support and improve highly sensible—if not highly critical—business decisions.
The term “intelligent businesses” does notmerely refer to commercialcompanies
but also to (hopefully)intelligent governments,intelligently managed educational
institutions,efficienthospitals,andsoon.Everycomplexbusinessactivitycanprofit
fromBI.
BIhasbecomeamainstreamtechnologyandis—accordingtomostinformation
technology analysts—looking forward to a more brilliant and prosperous future.
Almostallmediumandlarge-sizedenterprisesandorganizationsareeitheralready
using BI software or plan to make use of it in the next few years. There is thus a
rapidly growing need of BI specialists. The need of experts in machine learning
and data analytics is notorious. Because these disciplines are central to the Big
Data hype,and because Google,Facebook,and other companiesseem to offeran
infinite numberof jobs in these areas, students resolutely require more courses in
machine learning and data analytics. Many Computer Science Departments have
consequentlystrengthenedtheircurriculawithrespecttotheseareas.
However, machine learning, including data analytics, is only one part of BI
technology.Beforea “machine”canlearnfromdata,oneactuallyneedsto collect
thedataandpresenttheminaunifiedform,aprocessthatisoftenreferredtoasdata
provisioning.This, in turn, requiresextracting the data from the relevantbusiness
processes and possibly also from Web sources such as social networks, cleaning,
transforming,andintegratingthem,andloadingthemintoadatawarehouseorother
typeof database.To make humansefficientlyinteractwith variousstagesof these
activities,methodsandtoolsfordatavisualizationarenecessary.BIgoes,moreover,
much beyond plain data and aims to identify, model, and optimize the business
processesofanenterprise.AlltheseBIactivitieshavebeenthoroughlyinvestigated,
andeachhasgivenrisetoanumberofmonographsandtextbooks.Whatwassorely
missing,however,wasa bookthattiesitalltogetherandthatgivesa unifiedview
ofthevariousfacetsofBusinessIntelligence.
v
vi Foreword
The present book by Wilfried Grossmann and Stefanie Rinderle-Ma brilliantly
fills this gap. This book is a thoughtfulintroduction to the major relevant aspects
of BI. The book is, however, not merely an entry point to the field. It develops
the various subdisciplines of BI with the appropriate depth and covers the major
methods and techniques in sufficient detail so as to enable the reader to apply
themin areal-worldbusinesscontext.Thebookfocuses,inparticular,onthefour
major areas related to BI: (1) data modelingand data provisioningincludingdata
extraction, integration, and warehousing; (2) data and process visualization; (3)
machinelearning,dataandtextmining,anddataanalytics;and(4)processanalysis,
mining,andmanagement.ThebookdoesnotonlycoverthestandardaspectsofBI
butalsotopicsofmorerecentrelevancesuchassocialnetworkanalyticsandtopics
ofmorespecializedinterestsuchastextmining.Theauthorshavedoneanexcellent
jobinselectingandcombiningalltopicsrelevanttoamodernapproachtoBusiness
Intelligenceandtopresentthecorrespondingconceptsandmethodswithinaunified
framework.To the best of my knowledge,this is the first bookthatpresents BI at
thislevelofbreadth,depth,andcoherence.
The authors, Wilfried Grossmann and Stefanie Rinderle-Ma, joined to form
an ideal team towards writing such a useful and comprehensive book about
BI. They are both professors at the University of Vienna but have in addition
gainedsubstantialexperiencewithcorporateandinstitutionalBIprojects:Stefanie
Rinderle-MamoreintheprocessmanagementareaandWilfriedGrossmannmore
inthefieldofdataanalytics.Totheprofitofthereader,theyputtheirknowledgeand
experiencetogether to develop a common languageand a unified approachto BI.
Theyare,moreover,expertsinpresentingmaterialtostudentsandhaveatthesame
time the real-life background necessary for selecting the truly relevant material.
Theywereabletocomeupwithappropriateandmeaningfulexamplestoillustrate
themainconceptsandmethods.Infact,thefourrunningexamplesinthisbookare
groundedinbothauthors’richprojectexperience.
ThisbookissuitableforgraduatecoursesinaComputerScienceorInformation
Systemscurriculum.Atthesametime,itwillbemostvaluabletodataorsoftware
engineerswhoaimatlearningaboutBI,inordertogaintheabilitytosuccessfully
deployBItechniquesinanenterpriseorotherbusinessenvironment.Icongratulate
theauthorsonthiswell-written,timely,andveryusefulbook,andIhopethereader
enjoysitandprofitsfromitasmuchaspossible.
Oxford,UK GeorgGottlob
March2015
Preface
The main task of business intelligence (BI) is providing decision support for
businessactivitiesbasedonempiricalinformation.Thetermbusinessisunderstood
in a rather broad sense covering activities in different domain applications, for
example, an enterprise, a university, or a hospital. In the context of the business
under consideration, decision support can be at different levels ranging from the
operationalsupportforaspecificbusinessactivityuptostrategicsupportatthetop
levelofanorganization.Consequently,thetermBIsummarizesahugesetofmodels
and analytical methodssuch as reporting,data warehousing,data mining, process
mining,predictiveanalytics,organizationalmining,ortextmining.
In this book, we present fundamental ideas for a unified approach towards BI
activitieswithanemphasisonanalyticalmethodsdevelopedintheareasofprocess
analysisandbusinessanalytics.
ThegeneralframeworkisdevelopedinChap.1,whichalsogivesanoverviewon
thestructureofthebook.Oneunderlyingideaisthatallkindsofbusinessactivities
areunderstoodasaprocessintimeandtheanalysisofthisprocesscanemphasize
different perspectives of the process. Three perspectives are distinguished: (1)
the production perspective, which relates to the supplier of the business; (2) the
customerperspective,whichrelatestousers/consumersoftheofferedbusiness;and
(3)theorganizationalperspective,whichconsidersissuessuchasoperationsinthe
productionperspectiveorsocialnetworksinthecustomerperspective.
Core elements of BI are data about the business, which refer either to the
description of the process or to instances of the process. These data may take
different views on the process defined by the following structural characteristics:
(1) an event view, which records detailed documentation of certain events; (2) a
stateview,whichmonitorsthedevelopmentofcertainattributesofprocessinstances
over time; and (3) a cross-sectional view, which gives summary information of
characteristic attributes for process instances recorded within a certain period of
time.
The issues for which decision support is needed are often related to so-called
keyperformanceindicators(KPIs)andtotheunderstandingofhowtheydependon
certaininfluentialfactors,i.e.,specificitiesofthebusiness.Foranalyticalpurposes,
vii
viii Preface
it is necessary to reformulate a KPI in a number of analytical goals. These goals
correspond to well-known methods of analysis and can be summarized under
the headings business description goals, business prediction goals, and business
understandinggoals.Typicalbusinessdescriptiongoalsarereporting,segmentation
(unsupervisedlearning),andtheidentificationofinterestingbehavior.Businesspre-
dictiongoalsencompassestimationandclassificationandareknownassupervised
learninginthe contextofmachinelearning.Businessunderstandinggoalssupport
stakeholdersin understandingtheir business processesandmay consistin process
identificationandprocessanalysis.
Basedonthisframework,wedevelopamethodformatforBIactivitiesoriented
towardsideasoftheL(cid:2)formatforprocessminingandCRISPforbusinessanalytics.
Themaintasksoftheformatarethebusinessanddataunderstandingtask,thedata
task, the modeling task, the analysis task, and the evaluation and reporting task.
Thesetasksdefinethestructureofthefollowingchapters.
Chapter 2 deals with questions of modeling. A broad range of models occur
in BI corresponding to the different business perspectives, a number of possible
views on the processes, and manifold analysis goals. Starting from possible ways
of understandingthe term model,the mostfrequentlyused modelstructuresin BI
are identified, such as logic-algebraic structures, graph structures, and probabilis-
tic/statisticalstructures.Eachstructureisdescribedintermsofitsbasic properties
and notation as well as algorithmic techniques for solving questions within these
structures.Backgroundknowledgeisassumedaboutthesestructuresatthelevelof
introductorycoursesinprogramsforappliedcomputerscience.Additionally,basic
considerations about data generation, data quality, and handling temporal aspects
arepresented.
Chapter3elaboratesonthedataprovisioningprocess,rangingfromdatacollec-
tionandextractiontoasoliddescriptionofconceptsandmethodsfortransforming
dataintoanalyticaldataformatsnecessaryforusingthedataasinputforthemodels
in the analysis. The analytical data formats also cover temporal data as used in
processanalysis.
InChap.4,wepresentbasicmethodsfordatadescriptionanddatavisualization
thatareusedinthebusinessanddataunderstandingtaskaswellasintheevaluation
andreportingtask. Methodsfor process-orienteddata and cross-sectionaldata are
considered.Basedonthesefundamentaltechniques,wesketchaspectsofinteractive
anddynamicvisualizationandreporting.
Chapters5–8explaindifferentanalyticaltechniquesused forthe main analysis
goals of supervised learning (prediction and classification), unsupervised learning
(clustering), as well as process identification and process analysis. Each chapter
is organized in such a way that we first present first an overview of the used
terminologyandgeneralmethodologicalconsiderations.Thereafter,frequentlyused
analyticaltechniquesarediscussed.
Chapter 5 is devoted to analysis techniques for cross-sectional data, basically
traditional data mining techniques. For prediction, different regression techniques
are presented. For classification, we consider techniques based on statistical prin-
ciples, techniques based on trees, and support vector machines. For unsupervised
Preface ix
learning, we consider hierarchical clustering, partitioning methods, and model-
basedclustering.
Chapter 6 focuses on analysis techniques for data with temporal structure. We
startwithprobabilistic-orientedmodelsinparticular,Markovchainsandregression-
based techniques (event history analysis). The remainder of the chapter considers
analysis techniques useful for detecting interesting behavior in processes such as
associationanalysis,sequencemining,andepisodemining.
Chapter 7 treats methods for process identification, process performance man-
agement, process mining, and process compliance. In Chap. 8, various analysis
techniques for problems are elaborated, which look at a business process from
differentperspectives.Thebasicsofsocialnetworkanalysis,organizationalmining,
decision point analysis, and text mining are presented. The analysis of these
problemscombinestechniquesfromthepreviouschapters.
For explanationof a method, we use demonstrationexamples on the one hand
andmorerealisticexamplesbasedonusecasesontheotherhand.Thelatterinclude
the areas of medical applications, higher education, and customer relationship
management. These use cases are introduced in Chap. 1. For software solutions,
wefocusonopensourcesoftware,mainlyRforcross-sectionalanalysisandProM
forprocessanalysis.Adetailedcodeforthesolutionstogetherwithinstructionson
howtoinstallthesoftwarecanbefoundontheaccompanyingwebsite:
www.businessintelligence-fundamentals.com
The presentation tries to avoid too much mathematical formalism. For the
derivation of properties of various algorithms, we refer to the corresponding
literature. Throughout the text, you will find different types of boxes. Light grey
boxesareusedforthepresentationoftheusecases,darkgreyboxesfortemplates
thatoutlinethemainactivitiesinthedifferenttasks,andwhiteboxesforoverview
summariesofimportantfactsandbasicstructuresofprocedures.
The materialpresentedin the bookwas used by the authorsin a 4-h course on
Business Intelligence running for two semesters. In case of shorter courses, one
couldstartwithChaps.1and2,followedbyselectedtopicsofChaps.3,5,and7.
Vienna,Austria WilfriedGrossmann
Vienna,Austria StefanieRinderle-Ma