De presentatie van Eric van Heck, tijdens de parallelle sessie 'Methoden en technieken voor data-analyse' van het congres 'Data gedreven Beleidsontwikkeling' in Den Haag op 28 november 2017.
Eric van Heck - Congres 'Data gedreven Beleidsontwikkeling'
1. CHALLENGES WITH
BUSINESS ANALYTICS
RESEARCH:
TURNING DATA INTO BUSINESS
DEPARTMENT OF TECHNOLOGY & OPERATIONS MANAGEMENT
PROF. DR. IR. ERIC VAN HECK
CHAIRMAN DEPARTMENT OF TECHNOLOGY & OPERATIONS MANAGEMENT
DATA DRIVEN POLICY CONFERENCE, THE HAGUE, 28 NOVEMBER, 2017
3. TURNING DATA INTO BUSINESS
How to navigate in
a digitizing world?
• Strategy
• People
• Data & Information
• Business Models
• Digital Technology
• Transformation
• Society
4. TURNING DATA INTO BUSINESS
Business
Goal
Sensing
Data Storage
Analysing
Responding
Learning
8. Cascais Data Science for Social Good Europe
Summer Fellowship 2017
dssg.uchicago.edu/europe @DSSG_Europe
Identifying Green Rooftops in Rotterdam to
Improve Urban Planning
The Municipality of Rotterdam
Rodrigo
Belo
Tiago
Louro Alves
Diogo
Conceição
Carlos
Gonçalves
N. Gizem
Bacaksizlar
João Dinis
Fernandes
João
Louro
Qiwei
Han
9. Extension of the
public space
at ground level
Solar
Panels
Green Roof
Leisure
Reservoir
How can we help Rotterdam identify its
green rooftops efficiently?
16. EXPECTED VALUE FRAMEWORK: ALTERNATIVE
FORMULATION USING THE CONFUSION MATRIX
• The Expected Value Framework now becomes:
• EV = p(Y,p).b(Y,p) + p(N,p).b(N,p)+ p(Y,n).b(Y,n) + p(N,n).b(N,n)
p n
Model: Y b (Y,p) b(Y,n)
Model: N b(N,p) b(N,n)
Expected Value of true positivesExpected Value of false negativesExpected Value of false positivesExpected Value of
true negatives
• b(.) is the benefit (costs are negative
benefits) of each of the four possible
outcomes in a confusion matrix
– b(Y,p) = benefit of a true positive
– b(N,p) = benefit of a false negative
– b(Y,n) = benefit of a false positive
– b(N,n) = benefit of a true negative
(Provost & Fawcett, 2013; and Belo and Koppius, 2016)
17. EXPECTED VALUE FRAMEWORK IN THE CONFUSION
MATRIX FORMULATION HAS THREE IMPORTANT BENEFITS
1. Formulating the benefits of true positives, false
negatives etc. (so the b(Y,p), b(N,p), etc.) is often
a much easier way to incorporate the business
knowledge regarding value than the value of each
outcome
2. It allows for a better way of comparing different
models, especially when classes are unbalanced
3. It offers a framework for analyzing investments in
data
18. WHAT IS THE VALUE FOR ME?
2. Business Effectivity
big data and customer analytics
1. Business Efficiency
big data and roof top analytics
21. 21
Auctioneer’s Problem
Auctioneer:
• Clock speed?
• Starting price?
• Reserve price?
• Minimum purchase quantity?
• Revelation policies?
• Whether to bid
• At what time/price
• How many to buy
• Whether to bid
• At what time/price
• How many to buy
• Whether to bid
• At what time/price
• How many to buy
A Stylized Example
23. Comparison of Estimated and True Probability
Distribution Functions with Monte-Carlo Simulation
Four States:
{s00,s01,s10,s11}
Supply is low
versus high
Bidder i has
not won
versus
purchased 1
unit in
previous
rounds
24. THREE PROJECTS TO ADVICE THE
AUCTIONEER
1. What is the best minimum purchase quantity?
• Approach: structural modelling and Monte Carlo simulation based
on previous bids to determine the bidder’s value function.
• Result: Advice for minimum transaction amount policy
2. What type of bidders are there and what is their impact?
• Approach: K-means clustering based on previous bids and
multinomial logistic regression (MNL) to explain the bidder strategy.
• Result: five type of bidders with each a specific entry/exit and
bidding strategy and its impact on expected auction prices.
3. What is the impact of showing the winning bidder_ID?
• Approach: lab and field experiments and Difference-in-Differences
(DID) analysis of field experimental data (including a control group).
• Result: bidder_ID removed from most auction clocks.
25. WHAT IS THE VALUE FOR ME?
2. Business Effectivity
big data and customer analytics
3. New Business:
e.g. using your electric vehicles for energy
storage and balancing in the grid
1. Business Efficiency
big data and roof top analytics
26. 26
Micha Kahlen & Wolf Ketter
Erasmus Centre for Future Energy Business
kahlen@rsm.nl
27. ELECTRIC VEHICLES (EV’S) FROM CAR2GO
IN SAN DIEGO, USA
FleetPower committed strategically placed EV's as virtual power
plants to charge or discharge (Vehicle-2-Grid)
San Diego Amsterdam Stuttgart
Battery
technology 2015
(0.1 $/kWh)
3.7%
Gross profit increase
3.4%
Gross profit increase
4.4%
Gross profit increase
Battery
technology 2020
(0.05 $/kWh)
4.1%
Gross profit increase
3.9%
Gross profit increase
5.0%
Gross profit increase
Battery
technology 2025
(0.025 $/kWh)
4.8%
Gross profit increase
4.4%
Gross profit increase
6.1%
Gross profit increase
27
29. CONTRIBUTIONS
• Create sustainable new revenue streams for
electric vehicle rental companies without
compromising their rental business (customer
inconvenience)
• Limitations to profitability: regulation market
price, battery cost, infrastructure
• Enable the adoption of large scale renewable
energy sources
33. CHALLENGES
1. There Is No Common Digitized Platform
See also, for example, research by Davenport (2012),
Weill and Ross (2009), Westermand et a. (2014),
Wixom and Beath (2014).
36. CHALLENGES
1. There Is No Common Digitized Platform
2. Users Will Not / Can Not Engage
3. Lack of Commitment to Transform
37. TOP MANAGEMENT INVOLVEMENT
• Does your company board knows about:
– The potential value of big data & analytics
– Digitized platforms and the role of API’s
(Application Programming Interfaces)
– The potential role of Internet of Things (IoT)
technology
– The user interface of the “Uber” mobile app.
38. CHALLENGES
1. There Is No Common Digitized Platform
2. Users Will Not / Can Not Engage
3. Lack of Commitment to Transform
4. Digital Masters as New Entrants
39. Thank You
Please contact:
Erasmus Center for
Data Science and Business Analytics
Dr. Marcel van Oosterhout
Email: moosterhout@rsm.nl