IBM recently launched an updated version of its predictive analytics platform. Explore the latest features, including R, Python and Spark integration and more powerful decision optimization.
Fuel for the cognitive age: What's new in IBM predictive analytics
1. What’s new in IBM
Predictive Analytics
Fuel for the cognitive age:
2. It’s the dawn of the cognitive age
Where digital intelligence meets digital business
A cognitive business:
Thinks and learns from
data.
Takes many different
approaches to data.
This session covers a range of approaches to data to help you not only
think but outthink traditional analytical methods and identify new
opportunities.
3. Analytics-driven organizations reap rewards
Business outcomes
(69%)
Revenues
(60%)
Competitive
advantage
(53%)
Front Runners outperform on…
Source: IBM Institute for Business Value
(IBV)
…by using data and analytics
5. IBM Advanced Analytics Today
Data
Preparation
Analytics
at Scale
Insight to
Action
ALL Data
(Structured, Unstructured,
Streaming)
All Decisions
(People, Systems,
Strategic, Operational,
Real-Time)
• Predictive models
• Machine Learning
• Statistical Analysis
• Decision Optimization
• Real-Time Scoring
• Optimized Decisions
• APIs & services
• Dashboards /
Interactive apps
• Data models
• Data connectors
• Data Wrangling
Analytic
ServerModeler CPLEX StudioStatistics
IBM® (SPSS®) Predictive Analytics IBM® Prescriptive Analytics
Decision Optimization
Center (DOC)
DOCloud
6. Torchbearing CEOs look to predictive analytics in a
changing business landscape
More digital interaction
More competition expected
from other industries
82%
60%
60%
40%
Business landscape changes (in 3 to 5 years)
+37%
+50%
2015
2013
2015
2013
Insights from IBM’s Global C-suite Study – The CEO Perspective
ibm.com/csuitestudy
66%
50%
Use Predictive Analytics
Market Follower CEOsTorchbearer CEOs
more
32%
7. Predictive analytics can use virtually any data to
improve virtually any decision
Donor
management
Fraud
detection
Student
success
Sales
forecasting
Employee
turnover
Insurance
claims fraud
Cross-sell
and upsell
Customer
retention
8. Oak Lawn Marketing, Inc. employs a predictive analytics solution to
understand customer buying patterns and to target infomercials
Fourfold increase
in total revenue expected over a
three-year period as a result of
infomercials and other campaigns
Targets marketing
messaging and campaigns to
enhance the customer experience
and encourage retention
159% boost
in the average monthly rate of
customers who return to shop
compared to the previous year
Solution components
• IBM® SPSS® Modeler
• IBM Training
• IBM Business Partner AIT
Business challenge: Although Oak Lawn Marketing, Inc. was
gathering and generating enormous amounts of data about its
programs, the company couldn’t conduct a thorough analysis of
customers’ buying patterns using its outdated business intelligence
tools or unwieldy spreadsheets. Oak Lawn Marketing needed a
predictive analytics solution that would accurately portray and predict
customers’ buying trends and help it drive marketing campaigns.
The smarter solution: The solution combines predictive analytics,
rules, scoring and modeling algorithms as it analyzes transactional
and demographics information to help Oak Lawn Marketing
understand which products customers are most likely to purchase
and to guide the company in its decision-making processes. Using
this information, the company can customize its multitude of
infomercials with messaging appropriate for various TV channels
and time slots and tailor other marketing campaigns, such as
Internet and direct mail.
“We want to establish a brand that is used by our customers over a
long period of time.”
—Harry Hill, president and chief executive officer
9. • IBM® PureData™ System for
Analytics (powered by Netezza®
technology)
• IBM SPSS® Collaboration and
Deployment Services
• IBM SPSS Modeler
• IBM SPSS Modeler Desktop
• IBM SPSS Modeler Server
• IBM Training: SPSS
80% reduction
in serious accidents among
trucking company customers
Solution Components
Business Challenge: To meet customers’ demands, FleetRisk
Advisors needed to extract even deeper predictive insights regarding
truck driver safety from an ever-growing range of measured
parameters and get it done faster so that customers would have the
time to take truly preventive action.
The Smarter Solution: For each of the company’s customers’ truck
drivers, a powerful new predictive modeling solution translates some
4,500 data elements, from a diverse and ever-growing range of
sources, into quantitative risk ratings related to the likelihood of on-
the-job accidents, giving operators the cue they need to intervene
proactively to prevent such accidents and to save lives.
“Our new solution has enabled us to push the boundaries of predictive
risk analysis, which has translated into real value for our trucking
operator customers that rely on it.”
—Patrick Ritto, chief technology officer
20% reduction
in the incidence of minor
accidents
30% increase
in driver retention rates, with
commensurate decreases in
recruiting and training costs
FleetRisk Advisors helps trucking operators prevent more accidents
by building stronger and faster risk prediction models
10. What’s new in IBM SPSS Predictive Analytics
Empower
every user
Unlock
more data, faster
Ground to Cloud
deployment options
Code optional, open
to open source
Big Data for the
desktop
Predictive
everywhere
12. Uncover the Value of the Silent Data Majority
80% of data is unstructured; therefore, invisible
to computers and of limited use to business.
By 2020, 1.7MB of new information will be
created every minute for every human being
on the planet.
Incorporate Geospatial Data
Text Analytics with
Sentiment Analysis
Entity Analytics
Massively Parallel Algorithms
…delivered to your desktop!
13. Simplicity…
Without Sacrifice
Automatic data preparation
Automated model
creation
Infinitely complex
workflows
Advanced capabilities (text
analytics, entity analytics,
scripting)
14. Code-free deployment at scale: Activating analytics
Parallel In-Database
Optimized for Big Data environments
Reduce network traffic
Improved processing speed
Reduce data movement SQL pushback
Optimize performance with in-database
adapters
Increase analytic flexibility with in-database
mining
Advanced Model Management
(including A/B Testing, Champion/Challenger)
In-Database/In-Hadoop
Batch/Real-Time/Streaming
Point of Impact
(Analytical Decision Management)
15. Ground to cloud: Deployment flexibility
• Full breadth of analytical
capabilities
• Collaboration and
enterprise-wide best
practices
• Customize for (virtually)
any use case
• On-premises, hybrid and
software-as-a-service
LOB & Personal
Analytics
Developer
Tools
• Analytic tools built for
business
• Digitally delivered, digitally
fulfilled
• Variety of licensing and
packaging options
• Available for Windows or
Mac
• Build smarter data
applications -- quicker
• Endless possibilities
• No installation, no
configuration
• Mix & match components
Enterprise
Analytics
IBM SPSS Modeler
Gold
IBM Cloud Marketplace
(SPSS & DOCloud)
IBM Predictive
Analytics on Bluemix
16. Predictive Analytics on Bluemix
https://www.ng.bluemix.net/docs/#services/PredictiveModeling/index.html#pm_service
17. IBM SPSS & Decision Optimization in marketplace
https://www.ibm.com/marketplace/cloud/u
s/en-us
18. Open source and IBM Predictive Analytics
First R, then Python,
now Spark
Make coding optional
FacilitateEmbrace Extend
Make it massively
useful
19. Extend capabilities through open source: R
R Integration
R Build/Score, Process and Output node support
Scale R execution by leveraging database vendor
provided R engines
Custom Dialog Builder for R
Provides the ability to create new
Modeler Algorithm nodes and dialogs
that run R processes
Makes R usable for non-programmers
20. Growing catalog of extensions
IBM SPSS Community:
developer.ibm.com/predictiveanalytics
21. NEW! Python for Spark
Data Scientists can create
extensions for novice users to
exploit R, MLlib algorithms and
other Python processes
− Spark & its machine learning
library (MLlib)
− Other common Python libraries
• e.g.: Numpy, Scipy, Scikit-
learn, Pandas
Abstracting code behind a GUI
makes Spark usable for non-
programmers
23. IBM Decision Optimization for Python
Model & solve optimization
problems writing pure Python
Notebook ready.
Community-based
documentation and samples
Solving on cloud or
locally is invisible to APIs*
Access the
Technology Preview at :
pypi : docplex
Github : docplex
IBM market leading Decision
Optimization technology,
CPLEX, is now accessible as a
Pure Python package
under the
Apache license
(*) Solves the same program for free with IBM DOcloud free trial subscription or
IBM CPLEX Optimization Studio Community Edition installation
24. Beyond predictive
Capture price,
product,
location and
date for each
transaction.
Historical
& Master
Data ETL
Determine
important
variables,
predict trends,
seasonality etc.
Predictions
and Insights
Allow multiple
users to
experiment with
multiple
scenarios.
Collaboration
& What-if
Set policies,
promotions etc.
Allow reviewers
and auditors to
have a say.
Rules &
Process
Management
Automatically
generate
decisions, allow
user interaction
with decisions.
Decision
Making
Key steps for a mature decision support application leveraging
advanced analytics
Descriptive
Predictive
Prescriptive
25. Optimization is about resource
efficiency/utilization and allocation
Resources Choices to make
Capital Invest, allocate
People Hire, assign, schedule
Equipment Acquire, schedule, locate
Facilities Locate, size, schedule, maintain
Vehicles
Acquire, route, schedule, deliver,
maintain
Material/Product
Acquire, allocate, produce, deliver,
maintain
Keywords:
• minimize, maximize,
• how many/how much, which, when/where
• decide/choose, plan, schedule, assign, route, source, maintain,
locate, trade-off
Planning and scheduling activities
– Which are subject to complex
operating constraints (e.g.
limited resources, large volume
of data, complex manufacturing
or design processes)
– With multiple business objectives
to reduce time, cost, or increase
KPI’s such as productivity
While enabling
– Adjustment of changes in
operating environment
– What-if analysis
26. Optimization solutions – Documented ROI
2 Chilean Forestry firms Timber Harvesting $20M/yr + 30% fewer trucks
UPS Air Network Design $40M/yr + 10% fewer planes
South African Defense Force/Equip Planning $1.1B/yr
Motorola Procurement Management $100M-150M/yr
Samsung Electronics
Semiconductor
Manufacturing
50% reduction in cycle times
SNCF (French RR) Scheduling & Pricing $16M/yr rev + 2% lower op ex
Continental Airlines Crew Re-scheduling $40M/yr
AT&T Network Recovery 35% reduction spare capacity
Grantham Mayo van
Otterloo
Portfolio Optimization $4M/yr
Source: Edelman Finalists, http://www.informs.org or http://www.scienceofbetter.org
28. Optimization can utilize the full potential of SPSS
• Most planning considers only
averages
• However, advanced predictive
models create a range of
forecasts (lower and upper
bound / confidence interval)
• Instead of basing decisions on
a single forecast, you can use
the full range of information
provided by SPSS
29. Latest comparison
28
Start
• 12 instances
• 1st of every month
• 20 day horizon
Deterministic
• Single existing
forecast
Stochastic
• Create multiple
forecasts by
varying demand
forecast by +/-
10%
Deterministic
• Make decisions
based on single
forecast
Stochastic
• Make decisions
based on multiple
forecasts
Deterministic
• Compare with
actual demand
realization to
determine
revenue and cost
Stochastic
• Compare with
actual demand
realization to
determine
revenue and cost
30. Taking uncertainty into consideration
-0.1%-20%
improvement
~5.1% on average
~$23.9M annually
~$1.3M
on average
over 20 days