400 likes | 679 Views
Information and Service Management (ISM) 57E99902 Master’s thesis seminar. Merja Halme Timo Kuosmanen Pekka Malo . Statisticians ’ proverb. " Too much information kills the information and leads to misinformation ". Examples of Career Paths.
E N D
Information and Service Management (ISM) 57E99902 Master’sthesisseminar Merja Halme Timo Kuosmanen Pekka Malo
Statisticians’ proverb "Too much information kills the information and leads to misinformation"
Examples of Career Paths MattiKoivu, Executive at Financial Supervisory Authority MikkoSyrjänen, Executive at Wärtsilä Mari Aalto, Analyst at Customer Experience Unit, Itella MikkoSarkkinen, Consultant PirjaHeiskanen, Vice President, Trading and Industrial Intelligence, Fortum Veli-Pekka Heikkinen, Head of Portfolio Management, eQ Asset Management
Celebrities Terhi Majasalmi, appearing in tv series on personal budget management Sari Multala, world champion in sailing
Familiargraduates Jussi Pajunen, city mayor Sirpa Pietikäinen, MEP
Thesis titles from the past few years Media analytics: predicting returns with market sentiment Employing Data Mining in the Customer Relationship Management of a catalogue company Efficiency of retailers of Aro Corporation employing Data Envelopment Analysis
Thesis titles from the past few years Determining the ROI of a comprehensive CRM service - Case ItellaAsiakkuusmarkkinointi Menu planning of restaurants – integer optimization approach Use of self-organized maps in customer segmentation
Thesis titles from the past few years Fair allocation of costs in a joint project - Case HUS Subjective vs. objective methods in product cost forecasting Segmentation of Finnishgamblers Identification of fields of improvement in an internet-based gift store employing conjoint analysis
Getting started with your thesis … Data collection • Survey data • Company data (e.g. customer data) • Proprietary data banks • Streaming online data, etc. Analysis software • SPSS, SAS, R, RapidMiner, etc. Guidance for statistical inference
Questionnaires • Translate the info needed into questions • Mustmake the respondentinvolved and finish the interview • Minimize the responseerror Questionnaires
Questionnairemustbebased on previousresearch • Donot design a questionnairebeforecarefullycheckingwhathasbeenwritten on the topic, definingyourtheoreticalframework, studyinghowyoucanutilizeprevioussurveys (standardquestionsets, e.g. in marketing, psychology… ) • And getacquainted with questionnairetechniques Questionnaires
Importantpoints in questionnaireplanning 1/2 • What is being studied and why • What is the information needed • What is the inverviewing method • Content of individual questions • Question structure Questionnaires
Importantpoints in questionnaireplanning 2/2 • Questionwording • How to order the questions • Form and layout • Eliminateproblemsbypilot-testing • How to overcomeinability and unwillingness to answer Questionnaires
Whatkind of scales? The following is called the Likert scale. Which SCALE in our categorization? For me it is more important to aim at a job with a substantial salary than a job which is interesting. -3 Completely disagree • -2 Strongly disagree • -1 Somewhat disagree • 0 Neither disagree nor agree • 1 Somewhat agree • 2 Strongly agree • 3 Completely agree Questionnaires
Seeexamplesurvey • http://www.mmstudy.net/openg/openglogin.html • It is a good idea to useverydifferenttype of questions – REMEMBER TO ASK THE SAME THING REPEATEDLY FOR THE SAKE OF INCREASING RELIABILITY Questionnaires
Electronic questionnaire software • The studyused SSI Web • The schoolhasWebropol Questionnaires
Statistical software • Such as SAS EG or SAS Jmpor SPSS is the convenientway to do data analyses Database Marketing coursepagesinclude an introduction to SAS basicfunctions Questionnaires
One wayfrequencies • E.g. • How many responders have had more than one year of work experience? • Which percentage of responders is men? Questionnaires
Crosstabling E.g. • Create a cross table of gender and work experience? Does it seem that either males or females tend to have more working experience? • TESTING if the gender and working experience are independent/dependent Questionnaires
Comparingmeans • Calculate the means of WAGE for men and women and compare the Likert scale variable averages TESTING the equality of means?! Questionnaires
WE WILL HELP YOU • To makeyourownquestionnaires,interviews and analyses. The software is available and the websurveysareeasy to send and collect the data. Wewillassistyou with the analyses. • However, useyourtime to considerwhat to measure, how to ask, whatscales to use, how to takemissingvalues and how to achieve a goodsample. Questionnaires
Analysis of Public &Company Data • Typically the data usedearlierhasoriginatedfromcustomerdatabasesorithasbeenweb data (registeredusersorcookies) • Alsootherkinds of company data arepossible, e.g. in the numerousthesis made to assessefficiency of e.g. differentdepartments in a company • Data alsofrompublicsources, e.g. Statistics Finland
Company and Public Data – exampletoolsused • Econometricmodels • Forecasting • Efficiencyanalysis (e.g. DEA) • GameTheory • Optimization and Simulation • Data Mining (predictivemodeling, segmentation, association analysis etc.)
What is Multivariate Data Analysis? Multivariatestatistics ~ statisticalmethodsthatsimultaneouslyanalyseseveralmeasurements on eachdata-case (e.g. individual) underinvestigation. Analysingsituationswhenyouhave • Manyindependent (i.e. explanatory) variablesand/or • Manydependent (i.e. response/explained) variables • Varyingdegrees of correlationsbetweenvariables
WhyMultivariateStatistics? Difficulty of addressingcomplicatedresearchquestionswithunivariatetools Severaldrivers for increasingpopularity, e.g. • Availability of nicelypackaged software • Greatercomplexity of contemporaryresearch • Largeamounts of data • Emergence of data miningperspective (findingunforeseenpatterns and associations)
Quantitative Qualitative Confirmatory Qualitative & QuantitativeResearch Explorative
Hypothesis Population Data (Random Sample) Statistical Analysis Verification New Hypothesis / Theory Confirmatory Analysis Theory / Pre-Knowledge
Data Analysis Statistical Analysis New Theory / Invention Population New Data (Random) Statistical Analysis Verification New Hypothesis / Theory Exploratory Analysis Data (Sample)
Broad Topics Business Analytics Data mining Optimization Multi-criteria decision making Productivity and efficiency analysis
What is Data Mining? Discovery of knowledge from data in human-understandable form. Automatedextraction of predictiveinformation “Non-trivial extraction of implicit, previously unknown, and potentially useful information from data.” - Tan, Steinbach & Kumar : Introduction to Data Mining
From Data to Intelligent Business Decisions Source:Tan, Steinbach & Kumar: Introduction to data mining
Challenge of Big Data • DB • systems Artificial Intelligence Data Mining is a rapidlygrowingfieldwithin Business Intelligence Challenges: • highdimensional data • massivedatasets • variety of data types (numbers, images, text, etc.) • heterogenous and distributed data Data Mining is a marriage of manydisciplines
Common Data Mining Tasks Clustering Association Data (in different forms) Classification Regression Sequence analysis Outlier analysis
Applications and currentresearch Applications: marketanalysis, productrecommendation, shoppingbasketanalysis, economicforecasting and trendrecognition Currentresearchprojects: Financial sentimentanalysis Textmining for economictendencies Decisionmaking and performance in business simulations
Productivity & efficiencyanalysis Typicalresearchquestions: How organizationscouldutilizetheirresourcesmoreefficiently? Whichunitsperformwell? How muchinefficientunitscouldsaveresourcesorincrease output? How performancehasevolvedovertime? Applications: education, health care, energy, finance, agriculture, transport, utilities etc. Related fields: economics, management science, operations research, public administration.
Productivity & efficiencyanalysis Currentresearch at ourgroup: Stochasticnonparametricenvelopment of data (StoNED) • Modelingrisk and heteroscedasticity • Jointproduction of multipleoutputs • Non-convexfrontiers • Computationalalgorithms Currentapplications: Regulation of electricitydistribution Performance of a bankbranchnetwork Green growth in agriculture (OECD project)
Quantitative Finance Related Topics Evaluation of investment projects; e.g., investments in IT, services, etc. How valuable is a firm? Mergers and acquisitions Accounting for uncertainty in strategy decisions: major risk analysis Use of options and other instruments for risk management; real options ALM (asset-liability management) in insurance and banking Optimization of an investment portfolio (real or financial) Behavioral decision making; e.g., consumer choice behavior, private investment behavior Emissions trade and environmental protection And many more …