Post on 26-Jun-2015
description
Four Analytics Walk Into a Bar …
David F. Rogers
Department of Operations, Business Analytics, and Information Systems
Carl H. Lindner College of Business
University of Cincinnati
CBIG November 15, 2012 2
Prof.Apply.Skeptic.Gadfly.Challenge.Create
BS Math/Business 1978 – Murray State RacersMBA Quantitative Methods 1980 – Murray StatePhD Mgmt. – Quant. Methods & Ops. Mgmt. 1986 –
Krannert School @ Purdue BoilermakersUC Bearcats Lindner College of Business 1985-on.
◦Optimization Modeling /Analysis º Stochastic Modeling◦ Intro. Bus. Analytics & OR º Statistics º
Clustering◦Case Studies in Business Analytics
Toyota DunnHumby LexisNexis Landrum&Brown 5th3rd
OMI FlaggsUSA Merchants Cold Storage Kroger GE Aircraft Engines Kenton Co. Airport Board – CVG …
Research – OR; Applied Optimization
CBIG November 15, 2012 3
Traditional O.R. – BIG DATA, Big Help!
“Life is the Art of Drawing Sufficient Conclusions From Insufficient Premises” Samuel Butler, English Composer, Novelist, & Satiric Author (1835 – 1902)
Encounter a Problem or Opportunity…◦Qualitative Analysis Based on Management’s
Experience and Judgment◦Quantitative Analysis Based on Data,
Models, Analysis, and InterpretationMake a Decision – Like Eating
Mushrooms – Some are Poisonous!
CBIG November 15 2012 4
Factual Data, Regardless of How BIG, Can’t Replace Informed Judgment…
We Know Where the Crime is, but…◦How do We Best Modify Officer Assignments?◦How do We Respond to Immediate Changes
in the Data?◦Still Need the Experienced(?) Captain.
Player’s Points Scored. Sounds Simple. But…◦Per Game? Per Minute?◦Why Scored? Was the Best Point Guard
Playing at the Time?◦Still Need the Experienced(?) Coach.
CBIG November 15 2012 5
BIG DATA, Bigger Problems?
Little Bit of Data Gone Awry can Damage Analysis.
BIG DATA Collected Similarly Can Exacerbate That!
P&G Outsourced Data Collection.◦Some Regrets About Losing Control of
That.◦In-House Collection Can Also Be
Problematic… Data Collection from Dial Tones.
CBIG November 15, 2012 6
Data vs. Intuition…
CBIG November 15, 2012 7
Four Analytics Walk Into a Bar
The Four Analytic Characters …◦D – Descriptive Analytics – What Did
Happen?◦I – Inquisitive Analytics – Why Did it
Happen?◦P – Predictive Analytics – What Will
Happen?◦P – Prescriptive Analytics – What Should We Do?
D, I, P, and P Sip and Imbibe from …
BIG DATA.How Well Do They Walk Out? Let’s
See….
CBIG November 15, 2012 8
D – Descriptive Analytics – What Happened?
Just Give Me the Facts Ma’am…◦Frequencies, Minimums, & Maximums◦Mean, Medians, Modes, & Percentiles◦Standard Deviations & Ranges◦Skewness & Kurtosis◦Covariance & Correlation◦Confidence Intervals◦Bar/Pie Chart, DotPlot, Histogram,
Ogive, Stem&Leaf, & CrossTabs◦Visually Supported Well is Quite
Insightful.◦Academics Love This Development!
CBIG November 15, 2012 9
D – Descriptive Analytics – What Happened?
D Walks Out of the Bar On Steroids! Like Johnny Fever from WKRP in Cincy.
This is Where BIG DATA Rocks.◦Computer Advances in Hardware
& Software Make it… Easier to Collect & Store Enormous
Amounts. Easier to Visualize & Present.
◦Decomposable.◦Basic Statistics are More
Understandable to the Masses.
CBIG November 15, 2012 10
Be Careful! – Popular Infographics
CBIG November 15, 2012 11
But Be VERY CAREFUL…Recording ErrorsEmployee SabotageComputer GlitchesJaded DataDirty LaundryIncomplete, Missing, Contradictory,
Confidential, and/or Ambiguous.Irrelevant Data: “There are Three Reasons
Why I Can’t Do That. The First is That We Have No Money. And the Other Two Don’t Matter.”
NYC Mayor Fiorello LaGuardia
CBIG November 15, 2012 12
I – Inquisitive Analytics – Why Did it Happen?
With Overwhelming BIG DATA, Some of these May Become Moot with Population Info. ◦Sampling◦Confidence Interval Estimation◦Hypothesis Testing◦ANOVA
Portion of I that Doesn’t Become Moot Walks Out of the Bar Neatly Tailored…
◦More Sample Data Readily Available◦Higher Confidence Levels for Results
CBIG November 15, 2012 13
P – Predictive Analytics – What Will Happen?
P also Walks Out of the Bar Neatly Tailored.◦Regression Analysis & Prediction◦Forecasting Models◦Conjoint Analysis◦More Data to Choose From for More
Various Model Choices.
CBIG November 15, 2012 14
P – Prescriptive Analytics – What Should We Do?
BIG DATA Can be Overwhelming & P Does Not Walk Out of the Bar!◦Optimization Routines Can Grind to a Halt.◦Linear Programming w/ Continuous Variables
is OK.◦Integer Linear Programming – Mission Control We Have a Problem!◦Integer Nonlinear –Whoaaaa!!!! We
are Often Grappling in the Dark!◦Challenges for Researchers
Better Algorithmic Methods Better Computer Hardware
CBIG November 15 2012 15
Optimization Analysis …Problem Size & Solution Difficulty was
Already Problematic Before BIG DATA Advent. After, It is More Pronounced…
Example – Duke Provided Data & Wants to Cluster Time Periods for Differential Pricing. Hour
1 2 3 … 24 1
Building 2 kWh … Usage 93
CBIG November 15, 2012 16
Smart Meter BIG DATAModel MPS Minimize ZMPS
Subject to
CBIG November 15, 2012 17
1-Minute – 1,440 Time Periods
With Smart Meters, BIG DATA is Available and Much Finer than per Hour. 86,400=1Day
Hour Half-Hour Quarter-Hour 10-Min.
CBIG November 15 2012 18
Simulation …AKA, “Anti-Statistics” …
◦Statistics – BIG DATA Summarized with Few Numbers.
◦Simulation – Few Input Nos. & Generates BIG DATA.Response to a Lack of BIG DATA – Generate it.BIG DATA Implications for Simulation …
◦More Accurate Input Parameters. Natural Increased Confidence Levels with BIG DATA. Better Detailed Databases from Which to Choose Parameters.
◦More Appropriate and Sophisticated Models. Data Visualization Revelations Appended to Simulation Logic. Simulation Models Needed for More Scenarios. Be Careful of Over-Stated Models – Variable Interactions May
Exist. Current Softwares does Not Consider This.
CBIG November 15, 2012 19
Hierarchical PlanningWhat Level of Data is Needed?
◦Strategic – Corporate Level◦Tactical – Regional Level◦Operational – Plant Level◦Aggregation/Disaggregation Methods
MIT Work …◦Hax and Meal, etc….
CBIG November 15, 2012 20
Formal Education is Needed!
2011 Study by McKinsey Global Institute Predicts a Shortfall of 140,000 to 190,000 “Deep Analytical Positions” in the United States by 2018.
CBIG November 15, 2012 21
U.C. Master of Science inBusiness Administration (MSBA)
Business Analytics Concentration◦ Statistics º Simulation º Optimization◦ Visual Basic, SAS, AMPL, GAMS, Arena, Matlab, …◦ Capstone Experience is an Individual Project.
Information Systems Concentration◦ Data Visualization º Business Intelligence Project Management◦ DataBase Design º Data Warehousing º Data Mining◦ Text Mining º Enterprise Resource Planning (ERP)◦ IBM SPSS Data Modeler, ERWin for Dimensional Modeling, SAP◦ Capstone Experience is a Co-Op with Industry.
Certificate in Business Analytics – Started Fall 2012-13 http://business.uc.edu/future-students/graduate.html PhD in Business Analytics and Information Systems Also Available. Late Afternoon/Evening Classes.
CBIG November 15, 2012 22
CBIG November 15, 2012 23
INFORMS Analytics Magazine http://www.analytics-magazine.org/
CBIG November 15, 2012 24
INFORMS CAP
CBIG November 15, 2012 25
INFORMS Analytics Section
CBIG November 15, 2012 26
INFORMS LocallyCincinnati/Dayton Chapter of INFORMS
◦Three+ Activities/Year Summer Picnic at West Chester, OH Autumn Speaker & Business Meeting Spring Arnoff Lecture & Business Meeting at UC Joining INFORMS? Please Join the Cin/Day Chapter
Also!
UC INFORMS Student Chapter◦We Want You to Come Speak to Our Students!◦Mostly MSBA-BA Students.◦Great for Meeting Them for Your Hiring
Purposes.
CBIG November 15, 2012 27
How Can We Work Together?
David.Rogers@UC.edu(513)556-7143
Thanks!!!