28. Matrix = Associations
Rose Navy Olive
Alice 0 +4 0
Bob 0 0 +2
Carol -1 0 -2
Dave +3 0 0
Things are associated
Like people to colors
Associations have strengths
Like preferences and
dislikes
Can quantify associations
Alice loves navy = +4,
Carol dislikes olive = -2
We don’t know all
associations
Many implicit zeroes
Source: Sean Owen(2012), Cloudera
29. In Terms of Few Features
▣ Can explain associations by appealing to
underlying features in common (e.g. “blue-ness”)
▣ Relatively few (one “blue-ness”, but many shades)
(Alice)
(Blue)
(Navy)
Source: Sean Owen(2012), Cloudera
30. Losing Information is Helpful
▣ When k (= features) is small, information is lost
▣ Factorization is approximate
(Alice appears to like blue-ish periwinkle too)
(Alice)
(Blue)
(Navy)
(Periwinkle)
Source: Sean Owen(2012), Cloudera
32. ALS Algorithm
• Optimizing X, Y simultaneously is non-convex, hard
• If X or Y are fixed, system of linear equations: convex, easy
• Initialize Y with random values
• Solve for X
• Fix X, solve for Y
• Repeat (“Alternating”)
X
YT
54. Internet as a mass media
“Half the money I spend on advertising is
wasted; the trouble is I don‘t know which half.”
-- John Wanamaker, ~ 1875a pioneer in marketing
…
58. Common Data Categories
▣ Persona
u Age, Gender, Birth date,
City, …
▣ Attributes
u Phone brand/model, location,
time, App, browser, banner…
▣ Behavior
u Click,
Conversion(Installation, Cart,
Purchase, …), Activation,
Payment…
67. Bid Landscape Forecasting
▣ Only reference market price by base price
▣ Imp, UU, Click, Conv.
▣ DSP, per campaign, per targeting criteria…
▣ the advertisers’ targeting profiles à the winning bid value
▣ Pacing
Source: Ying Cui, Ruofei Zhang, Wei Li, Jianchang Mao(2011), Bid Landscape Forecasting in Online Ad Exchange Marketplace, Yahoo! Labs
68. Traffic forecasting
▣ An impression on Jeremy Lin BBS post of MiuPTT
▣ Two product ads
u A: Linsanity T-Shirt
u B: Baseketball shoes
▣ Not optimized if only bid for highest price
u B bid higher than A
u Inventory A is much fewer than inventory B
70. Source: Ying Cui, Ruofei Zhang, Wei Li, Jianchang Mao(2011), Bid Landscape Forecasting in Online Ad Exchange Marketplace, Yahoo! Labs
*
*C
1
*
*C
2
*
B1
C1
*
B1
C2
* *
*
* *
*
* *
C1
* *
C2
*
B1
*
• Remove few-imp path for
not too sparse
• Easily to target all
Target Attribute
Bid Landscape Forecasting
-- Bid Star Tree Expansion
71. Bidding Price Calculation
Bidding Price = F (base price, CVR )• In the same campaign, the conversion value is
the same
= base price * φ • φ = CVR / avg CVR
φ = p(c|u, i) / Ej [ p(c|u, j) ]
Ej [ p(c|u, j) ] = Σj p(c|u, j) p(j) = p(c|u)
• I, j : inventory (on Web/App)
φ = p(c|u, i) / p(c|u)
φ = p(c|s, i) / p(c|s)
• All inventories in the same segment are the same,
• i could also be inventory cluster
72. Bidding Price Model Building
• Cold start
• Training feature of segment and inventory
respectively
• Do not train combined feature for preventing over
fitting on few training data
φ = p(c|s, i) / p(c|s)
• Cold start
• Training feature of segment
AUC: the area under the ROC curve(TP/FP)
Lift: target response divided by average response.
bid = BasePrice(s, a) * p(c|s, I, a) / p(c|s, a)
DSP cross-campaigns:
73. Bidding Price Calibration
▣ Forecasting
u Sampling learn
▣ On-line adjustment
u Feedback control Re-learn
▣ Loss reason
u Prior Probability Shift
n Budget, Freq. cap, …
u Competition
77. 4R: Reach, Richness, Representation, Range
Reach
Richness
High
High
(DAU)
(Behavioral data)
Range
( Affiliate of whole
context)
Representation
(Format Content)
78. Data Economy
Traditional - Internet Economy
HighREACH
RICHNESS
High
Low
Traditional
Economy
Internet Economy
(quality)
(quantity)
79. Reach: The Value Funnel
CPM campaign:
Revenue = N/1000 ⋅CPM
CPC campaign:
Revenue = N ⋅ CTR ⋅ CPC
CPA campaign:
Revenue = N ⋅ CTR ⋅
CVR⋅ CPA
UU Reach (DAU)
ARPU = Life-time Value
82. Richness
▣ Data Quality Richness
u Attr. vs. behavior
▣ Data Utilization Richness
u Call taxi (short vs. long route)
u Download times vs. Activation days
▣ Data Model Richness
88. Range
- Roger Martin
Rothman School of Management, Toronto
If only attach importance to quantify the business
model, it will not have the ability to find a potential
growth opportunities: The pursuit of quantifying
the biggest problem is that people ignore the
context of the behavior generated, detached from
the context of the event, and have not been
included in the model ignores variables
effectiveness.
93. 4R: Reach, Richness, Representation, Range
Reach
Richness
High
High
(DAU)
(Behavioral data)
Range
( Affiliate of whole
context)
Representation
(Format Content)
94. World, Model Theory
Credit: John F. Sowa
generalized statements,
proven scientifically with evidence
Simplified representation, helpful tool to
understand specific phenomena
106. Ask Alexa – A Real Demo
▣ Alexa, what time is it? What date is it?
▣ Alexa, sing a happy song. Can you beatbox?
▣ Alexa, Happy Birthday/Xmas? twinkle twinkle little star
▣ Alexa, tell me a joke. Do you have a boy friend?
▣ Alexa, play some music / play Classical Study Music / play some jazz
▣ Alexa, inspire me. / open ocean sounds.
▣ Alexa, what's in the news?
▣ Alexa, what's the weather like?
▣ Alexa, tell me about the movie [title].
▣ Alexa, what's 56 times 33?
▣ Alexa, what's the definition of [word]? how do you spell [word]?
▣ Alexa, Thank you.
https://www.cnet.com/how-to/the-
complete-list-of-alexa-commands/
109. Paradigm Shift(Reverse)
▣ Move
u Data à program
▣ Value
u Things à Product Service à Personal Service
u Value/revenue shift
u What if phone price is near its cost or free?
143. The Revolution of Big Data
DATA
Hypotheses
Statistical Analysis
BIG DATA
Hypotheses
Machine Learning
Data Mining
Machine-generated
Sampling, Multi-variant… All, Hyper space, …
Volume, Velocity, Variety, Veracity
Human-explainable
144. Models ßà Cases
Russ Merz, An Integrated Model of Media Satisfaction and Engagement: Theory, Empirical
Assessment and Managerial Implications, Journal of Applied Marketing Theory, Nov 2011
145. Models ßà Cases
Russ Merz, An Integrated Model of Media Satisfaction and Engagement: Theory, Empirical
Assessment and Managerial Implications, Journal of Applied Marketing Theory, Nov 2011
BIG DATA
Hypotheses
Machine Learning
Data Mining
Machine-generated
All, Hyper space, …
Volume, Velocity, Variety, Veracity
deductive inductive
Cases
Models
Models
Cases
149. ▣ “The Theory of Moral Sentiments” by Adam Smith
u Every man is, no doubt, by nature first and principally
recommended to his own care; and he is fitter to care of
himself than of every other person… (1759, 82)
▣
159. Know-What, Know-Why, Know-How
and Decision Making
Know-How
(Feasible?)
Prescriptive
Know-What
(Objective?)
Descriptive
Know-Why
(Scientific?)
Normative
Descriptive D-M:
How decisions are made?
Normative D-M:
How decisions should be made?
Prescriptive D-M:
How decisions could be made better?
RationalityBounded Rationality
H. Simon,
Administrative
Behavior AI
Src: JT Chiang, NTU MBA