rm world 2014: big data vs. classic data
DESCRIPTION
TRANSCRIPT
1 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
BigData Vs. “Classic” Data: Predictive Analytics in a Changing Data Landscape
Usama Fayyad, Ph.D.
Chief Data Officer – Barclays
Twitter: @usamaf
August 19, 2014
RapidMiner World
Boston, MA
2 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Outline• Big Data all around us• The CDO role• Some of the issues in BigData• Introduction to Data Mining and Predictive
Analytics Over BigData• Case studies • Summary and conclusions
3 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
What Matters in the Age of Analytics?
1.Being Able to exploit all the data that is available • not just what you've got available • what you can acquire and use to enhance your actions
2. Proliferating analytics throughout the organization• make every part of your business smarter• Actions and not just insights
3. Driving significant business value • embedding analytics into every area of your business can
significantly drive top line revenues and/or bottom line cost efficiencies
4 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Why Big Data?A new term, with associated “Data Scientist” positions:
• Big Data: is a mix of structured, semi-structured, and unstructured data:– Typically breaks barriers for traditional RDB storage
– Typically breaks limits of indexing by “rows”
– Typically requires intensive pre-processing before each query to extract “some structure” – usually using Map-Reduce type operations
• Above leads to “messy” situations with no standard recipes or architecture: hence the need for “data scientists” – conduct “Data Expeditions”
– Discovery and learning on the spot
5 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
The 4-V’s of “Big Data”
• Big Data is Characterized by the 3-V’s:
– Volume: larger than “normal” – challenging to load/process• Expensive to do ETL
• Expensive to figure out how to index and retrieve
• Multiple dimensions that are “key”
– Velocity: Rate of arrival poses real-time constraints on what are typically “batch ETL” operations
• If you fall behind catching up is extremely expensive (replicate very expensive systems)
• Must keep up with rate and service queries on-the-fly
– Variety: Mix of data types and varying degrees of structure• Non-standard schema
• Lots of BLOB’s and CLOB’s
• DB queries don’t know what to do with semi-structured and unstructured data.
6 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Male, age 32
Lives in SFLawyer
Searched on from London last week
Searched on:“Italian restaurantPalo Alto”
Checks Yahoo! Mail daily via PC & Phone
Has 25 IM Buddies, Moderates 3 Y! Groups, and hosts a 360 page viewed by 10k people
Searched on:“Hillary Clinton”
Clicked on Sony Plasma TV
SS ad
Registration Campaign Behavior Unknown
Spends 10 hour/week
On the internet Purchased Da Vinci Codefrom Amazon
“Classic” Data: e.g. Yahoo! User DNA
7 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Male, age 32
Lives in SFLawyer
Searched on from London last week
Searched on:“Italian restaurantPalo Alto”
Checks Yahoo! Mail daily via PC & Phone
Has 25 IM Buddies, Moderates 3 Y! Groups, and hosts a 360 page viewed by 10k people
Searched on:“Hillary Clinton”
Clicked on Sony Plasma TV
SS ad
Spends 10 hour/week
On the internet Purchased Da Vinci Code from Amazon
How Data Explodes: really big
Social Graph (FB)
Likes &
friends likes
Professional netwk
- reputation
Web searches on
this person,
hobbies, work,
locationMetaData on everything
Blogs, publications,
news, local papers,
job info, accidents
8 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
The Distinction between “Classic Data” and “Big Data” is fast disappearing
• Most real data sets nowadays come with a serious mix of semi-structured and unstructured components:– Images– Video– Text descriptions and news, blogs, etc…– User and customer commentary– Reactions on social media: e.g. Twitter is a mix of data
anyway
• Using standard transforms, entity extraction, and new generation tools to transform unstructured raw data into semi-structured analyzable data
9 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Text Data: The Big Driver
• We speak of “big data” and the “Variety” in 3-V’s• Reality: biggest driver of growth of Big Data has been
text data– Most work on analysis of “images” and “video” data has
really been reduced to analysis of surrounding text
Nowhere more so than on the internet
• Map-Reduce popularized by Google to address the problem of processing large amounts of text data: – Many operations with each being a simple operation but
done at large scale– Indexing a full copy of the web– Frequent re-indexing
10 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
IT Log & Security Forensics & Analytics
Automated Device DataAnalytics
Failure Analysis
Proactive Fixes
Product Planning
AdvertisingAnalytics
Segmentation
Recommendation
Social Media
Big DataWarehouse Analytics
Cost Reduction Ad Hoc Insight
Predictive Analytics
Hadoop + MPP + EDW
Find New Signal Predict Events
100% Capture
Big Data Applications and Uses
11 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
A few words on:The Chief Data Officer
Why are companies creating this position?
12 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Why a Chief Data Officer?• There is a fundamental realisation that Data needs to
become a primary value driver at organizations• We have lots of Data• We spend much on it: in technology and people• We are not realising the value we expect from it
• A strong business need to create the CDO role:• Traditional companies are not following, but adopting the
model that actually works in other data-intensive industries
• CDO has a seat at executive table: the voice of Data• Data done right is an essential element to unify large
enterprises to unlock value form business synergies
13 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
What about traditional IT?• Does your IT department provide you with the data you
need?• Or is it just another stumbling block to get at Data?
• Does your IT department understand what you need for analytics?
• Or is it just about ETL, SQL databases, and restricted access?
• Does your IT department have people who understand data modeling? Data Warehousing? BI?• Or is it just about capturing transactions into standard
normalized databases? • Is BI more than just a tool for MI reporting?
14 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Fundamental Data Principles to Support Analytics
Usama’s Obvious Data Axioms
15 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Data Axioms
1. Data gains value exponentially when integrated and coalesced.
– When fragmented: dramatic value loss takes place;– increased costs; – reduced utility/integrity; – and increased security risks
16 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Data Axioms
2. Fusing Data together from disparate/independent sources is difficult
to achieve and impossible to maintainHence only viable approach is:• Intercepting and documenting at the source• fusing at the source • controlling lifecycle and flow
17 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Data Axioms
3. Standardisation is essential• for sustained ability to integrate data sources
and hence growing value; • for simplifying down-stream systems and apps• For enforcing discipline as a firm increases its
data sources
18 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Data Axioms
4. Data governance and policy must be centralised
• needs to be enforced strongly else we slip into chaos and a Babylon of terms/languages
• An Enterprise Data Architecture spanning structured and unstructured data
19 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Data Axioms
5. Recency Matters data streaming in modelling and scoring
• Often, accuracy of prediction drops quickly with time (e.g. consumer shopping)
• Value of alerts drop exponentially with time…• Ability to trigger responses based on real-
time scoring critical• Streaming, real-time model updates, real-
time scoring
20 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Data Axioms
6. Data Infrastructure Needs: • rapid renewal & modernization: the pace of
change and development of technology are very rapid– Design for migration and infrastructure replacement
via abstraction layers that remove tech dependencies
• Encryption and Masking: Persisting unencrypted confidential and secret data (even within secure firewalls) is an invitation for problems and risks
21 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Data Axioms
7. Data is a primary competency and not a side-activity supporting other
processes• Hence specialized skills and know-how are a
must• Generalists will create a hopeless mess• Data is difficult: modelling, architecture, and
design to support analytics
22 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Reality Check:Brand/Reputation on-line
What are people saying about my brand on Social Media?
23 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
24 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Reality Check
Surely there are companies I can work with that can help me make this
practical?
25 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
26 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Reality Check
So what do technology people worry about these days?
27 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
To Hadoop or not to Hadoop?
when to use techniques requiring Map-Reduce and grid computing?• Typically organizations try to use Map-Reduce
for everything to do with Big Data– This is actually very inefficient and often irrational– Certain operations require specialized storage
• Updating segment memberships over large numbers of users
• Defining new segments on user or usage data
28 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Drivers of Hadoop in Large Enterprises
Cost of Storage• Fastest growing demand is more storage• Data in Data Warehouses have traditionally required
expensive storage technology:
–$100K per terabyte per year – cost of Teradata storage
– $2.5K per terabyte – much lower per year –cost of Hadoop on commodity storage
29 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
ERP Financial Data1%
Supply Chain Data2%
Sensor Data2%
Financial Trading Data4%
CRM Data4%
Science Data7%
Advertising Data10%
Social Data11%
Text and Language Data16%
IT Log Data19%
Content and Preference Data24%
Hadoop Use Cases by Data Type
30 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Analysis & Programming Software
PIG
HIPI
31 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
32 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Reality: If Storage is Biggest Driver of Hadoop Adoption; What is the next biggest?
ETL• Replaces expensive licenses• Much higher performance with lower infrastructure
costs (processors, memory)• Flexibility in changing schema and representation• Flexibility on taking on unstructured and semi-
structured data• Plus suite of really cool tools…
33 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Turning the three Vs of Big Data into ValueUnderstand context and content• What are appropriate actions?• Is it Ok to associate my brand with this content?• Is content sad?, happy?, serious?, informative?
Understand community sentiment• What is the emotion?• Is it negative or positive?• What is the health of my brand online?
Understand customer intent?• What is each individual trying to achieve?• Can we predict what to do next?• Critical in cross-sell, personalization, monetization,
advertising, etc…
34 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Many Business Uses of Predictive AnalyticsAnalytic technique Uses in business
Marketing and sales Identify potential customers; establish the
effectiveness of a campaign
Understanding customer behavior model churn, affinities, propensities, …
Web analytics & metrics model user preferences from data, collaborative filtering, targeting, etc.
Fraud detection Identify fraudulent transactions
Credit scoring Establish credit worthiness of a customer requesting a loan
Manufacturing process analysis Identify the causes of manufacturing problems
Portfolio trading optimize a portfolio of financial instruments by maximizing returns & minimizing risks
Healthcare Application fraud detection, cost optimization, detection of events like epidemics, etc...
Insurance fraudulent claim detection, risk assessment
Security and Surveillance intrusion detection, sensor data analysis, remote sensing, object/person detection, link analysis, etc...
Case Studies:
1. Context Analysis (unstructured data)
2. Yahoo! predictive modeling
36 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Understanding Context
37 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Reality Check
So who is the company we think is best at handling BigData?
38 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Biggest BigData in Advertising?
Understanding Context for Ads
39 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
The Display Ads Challenge Today
What Ad would you place here?
40 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
The Display Ads Challenge TodayDamaging to Brand?
41 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
The Display Ads Challenge Today
What Ad would you place here?
42 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
The Display Ads Challenge Today
Irrelevant and Damaging to Brand
Completely Irrelevant
43 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
NetSeer: Intent for Display
• Currently Processing 4 Billion Impressions per Day
44 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Problem: Hard to Understand User Intent
Contextual Ad served by Google What NetSeer Sees:
Case Studies:
1. Context Analysis (unstructured data)
2. Yahoo! Predictive Modeling
Yahoo! – One of Largest Destinations on the Web
80% of the U.S. Internet population uses Yahoo!
– Over 600 million users per month globally!
Global network of content, commerce, media, search and
access products
100+ properties including mail, TV, news, shopping, finance,
autos, travel, games, movies, health, etc.
25+ terabytes of data collected each day
• Representing 1000’s of cataloged consumer behaviors
More people visited
Yahoo! in the past
month than:
• Use coupons
• Vote
• Recycle
• Exercise regularly
• Have children
living at home
• Wear sunscreen
regularly
Sources: Mediamark Research, Spring 2004 and comScore Media Metrix, February 2005.
Data is used to develop content, consumer, category and campaign
insights for our key content partners and large advertisers
Yahoo! Big Data – A league of its
own…Terrabytes of Warehoused Data
25 49 94 100500
1,000
5,000
Amaz
on
Kore
a
Tele
com
AT&T
Y! L
iveS
tor
Y! P
anam
a
War
ehou
se
Wal
mar
t
Y! M
ain
war
ehou
se
GRAND CHALLENGE PROBLEMS OF DATA PROCESSING
TRAVEL, CREDIT CARD PROCESSING, STOCK EXCHANGE, RETAIL, INTERNET
Y! Data Challenge Exceeds others by 2 orders of magnitude
Millions of Events Processed Per Day
50 120 225
2,000
14,000
SABRE VISA NYSE YSM Y! Global
Behavioral Targeting (BT)
Search
Ad Clicks
Content
Search Clicks
BT
Targeting ads to
consumers whose recent
behaviors online indicate
which product category is
relevant to them
Male, age 32
Lives in SFLawyer
Searched on from London last week
Searched on:“Italian restaurantPalo Alto”
Checks Yahoo! Mail daily via PC & Phone
Has 25 IM Buddies, Moderates 3 Y! Groups, and hosts a 360 page viewed by 10k people
Searched on:“Hillary Clinton”
Clicked on Sony Plasma TV
SS ad
Registration Campaign Behavior Unknown
Spends 10 hour/week
On the internet Purchased Da Vinci Codefrom Amazon
Yahoo! User DNA
• On a per consumer basis: maintain a behavioral/interests profile and profitability (user value and LTV) metrics
How it works | Network + Interests +
Modelling
Analyze predictive patterns for purchase
cycles in over 100 product categories
In each category, build models to describe
behaviour most likely to lead to an ad
response (i.e. click).
Score each user for fit with every
category…daily.
Target ads to users who get highest
‘relevance’ scores in the targeting
categories
Varying Product Purchase CyclesMatch Users to the ModelsRewarding Good BehaviourIdentify Most Relevant Users
Recency Matters, So Does Intensity
Active now… …and with feeling
Differentiation | Category specific
modelling
time
inte
nsity s
core
time
inte
nsity
score
Inte
nse
Clic
k Z
on
e
Example 1: Category Automotive Example 2: Category Travel/Last Minute
Different models allow us to weight and determine intensity and recency
Alt Behaviour 1: 5 pages, 2 search keywords, 1 search click, 1 ad click Alt Behaviour 1: 5 pages, 2 search keywords, 1 search click, 1 ad click
Inte
nse
Clic
k Z
on
e
Differentiation | Category specific
modelling
time
inte
nsity s
core
Intense Click Zone
Example 1: Category Automotive
Different models allow us to weight and determine intensity and recency
with no further activity, decay takes effect
Alt Behaviour 1: 5 pages, 2 search keywords, 1 search click, 1 ad click
user is in the Intense Click Zone
Automobile Purchase Intender Example
A test ad-campaign with a major Euro automobile manufacturer Designed a test that served the same ad creative to test and control groups
on Yahoo
Success metric: performing specific actions on Jaguar website
Test results: 900% conversion lift vs. control group Purchase Intenders were 9 times more likely to configure a vehicle, request
a price quote or locate a dealer than consumers in the control group
~3x higher click through rates vs. control group
Mortgage Intender Example
We found:
1,900,000 people looking
for mortgage loans.
+122%
CTR Lift
Mortgages Home Loans Refinancing Ditech
Financing section in Real Estate
Mortgage Loans area in Finance
Real Estate section in Yellow Pages
+626%
Conv Lift
Example search terms qualified for this target:
Example Yahoo! Pages visited:
Source: Campaign Click thru Rate lift is determined by Yahoo! Internal
research. Conversion is the number of qualified leads from clicks over number of impressions served. Audience size represents the audience within this behavioral interest category that has the highest propensity to engage with a brand or product and to
click on an offer.Date: March 2006
Results from a client campaign on Yahoo!
NetworkExample: Mortgages
Experience summary at Yahoo!
• Dealing with one of the largest data sources (25
Terabyte per day)
• Behavioral Targeting business was grown from $20M
to > $400M in 3 years of investment!
• Yahoo! Specific? -- BigData critical to operations
– Ad targeting creates huge value
– Right teams to build technology (3 years of recruiting)
– Search is a BigData problem (but this has moved to
mainstream)
Lessons Learned
A lot more data than qualified talent
Finding talent in BigData is very difficult
Retaining talent in BigData is even harder
At Yahoo! we created central group that drove huge value to
company
Data people need to feel like they have critical mass
Makes it easier to attract the right people
Makes it easier to retain
Drive data efforts by business need, not by technology
priorities
Chief Data Officer role at Yahoo! – now popular
What About RapidMiner
And Big Data?
What does this all mean to RapidMiner and its use in
enterprise analytics?
RapidMiner’s Strengths
5959
• Open Source Community & Marketplace – Crowd-sourced innovation, quality assurance, market awareness.
• Fully-integrated Platform – Integrated, process-based business analytics platform with focus on predictive analytics.
• No Programming Required – Easy-to-use, low maintenance costs, standard platform for business analysts.
• Advanced Analytics at Every Scale – In-memory, in-database and in-Hadoop analytics offer best option for every size of database.
• Connectivity – More than 60 connectors (incl. SAP & Hadoop), allowing easy access to structured and unstructured data.
30,000+ Downloads per Month
CONFIDENTIAL
SELECT LIST OF RECIPIENT ORGANIZATIONS
6060
Government & DefensePharma & Healthcare
Consulting
Oil & Gas, Chemicals
Financial ServicesSoftware & Analytics
Retail
Manufacturing
Business Services
Consumer ProductsAerospace
Technology
Entertainment Academia
Leader in Advanced Analytics
61
Source: Gartner
Magic Quadrant for Advanced
Analytics Platforms
(February 2014)
Full report at:
http://www.rapidminer.com/gartner
2014
62
PayPal
Who > world leading
online payment services
provider
Solution > Customer
feedback and voice of the
customer analysis, churn
prediction and prevention,
text mining and sentiment
analysis
SmartSoft
Who > provider of
solutions for preventing
fraud, money laundering,
and risks in financial
institutions
Solution > Integration of
Rapid-I’s predictive
analytics engine into their
solutions for fraud
detection and fraud
prevention for the
financial and telecom
sectors
Select Customer Stories
SO WHAT IS THE
BIG DATA STORY?
So the data is naturally moving to
Hadoop...
Situation:
–The data is moving to Hadoop for Cost (storage) and Convenience (ETL) forces
–How do we get the value of predictive analytics to the data?
Rather than move the data out, move the analytics to
the data!
–Can we minimize the need for data movement?
–Data copies can become a management nightmare
–Analytics on a “Business As Usual” manner require convenience
64
Radoop – RapidMiner on HadoopOpportunity:
–Avoid expensive data movement
–Leverage convenient data transformation
–Thousands of data connectors, many over semi-structured and unstructured data
Why is this big news?
–Leverages a naturally occurring wave
–Analytics over a richer variety requires much more processing
–The energy placed on data extraction and loading moves to energy applied on actual analysis and modelling
65
BigData Analytics & ThreatsOld data analysis regimes are breaking down
– They cannot accommodate all the new data sources from on-line and mobile
– Any data set can be enriched with semi-structured and unstructured data
–News articles
–User commentary and feedback
–Contextual awareness:
–Semantics of content–Semantics of location
– Data-driven marketing is key to innovation/platform success
Big Picture on Big Data Analytics
Observations and
Concluding Remarks
Retaining New Yahoo! Mail Registrants
Often,
Simple is Very Powerful!
Integrating Mail and News
Data showed that users often check their mail
and news in the same session
–But no easy way to navigate to Y! News from Y! Mail
Mail users who also visit Y! News are 3X more
active on Yahoo
–Higher retention, repeat visits and time-spent on Yahoo
“In the news” Module on Mail Welcome Page
Increased retention on Mail for light users by 40%!
– Est. Incremental revenue of $16m a year on Y! Mail alone
Nordstrom: Queries with No
Matches
Julie Bornstein, Web Marketing Director
–What are my customers looking for and not finding?
–June 2002: queries for “belly button rings”
–returned no matches in store
–Why the sudden interest?
Nordstrom: Queries with No
Matches
Print Ad Campaign
Models happen to be sporting a
navel ring
Nordstrom does not sell navel
rings
What to do???
Concluding Remarks
The early days of
mass auto
production
Today’s Auto:
It just works!
No need to understand what happens
when you turn on ignition
Very complex inside, but all simplicity
on the outside
“Gotcha”s in Big Data
Discounting the need for near-real-time data
processing and analytics
– Evils of “batch” thinking
No grid story – go build separate infrastructure to
learn
– Expensive
– Bad price-performance as utilization insufficient
Data Mining & Predictive analytics teams expect data
in their own stores
– Good luck using the analytic tools on your own DW store
– Build a “replica” or “extract” DW
Scoring and integration of results is a whole separate
story
Data Platform Considerations
Eliminate the evils of data extracts/movement out of
the main data store
Do you need map-reduce over a grid?
– Can you set up instant-on grid (Hadoop) as needed for the map-reduce tasks/demand without moving data?
How do you integrate with analytic functions?
– scoring, data transformations
– plug-in 3rd party analytics scoring apps
– Advanced dashboard and scorecard support
Built-in integration with statistical packages? SAS, R,
data mining algorithms, advanced analytics algorithms
libraries
Can you avoid data movement and additional copies?
78 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
BigData Analytics for Organizations
• Key to competitive Intelligence:
– Understand context
– Understand intent
• Key to understanding consumer trends through social media analysis
– Brand issues
– Trend issues
– Anticipating the next shift
79 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Threats & Opportunities• Data world is changing, especially in on-line businesses
• Major shifts from relational DB to NoSQL, document-oriented stores
• Connecting new world to “old” world?– Convenience of execution – integration with data platforms
– Appropriateness of algorithms to BigData
– Unstructured data algorithms:
• Text, Semi-structured and Unstructured data
• Entity extraction a must
• Appropriate theory and probability distributions (power laws, fat tails)
• Sparse Data
– Model management and proper aging of models
– Getting to basics so we can decide what models to use:
• Understanding noise and distributions
• Data tours
80 RapidMiner World Keynote talk – Copyright Usama Fayyad © 2014
Usama Fayyad - [email protected][email protected]
Twitter – @Usamaf
+1-206-529-5123
www.Oasis500.com
www.open-insights.com
Thank You! & Questions