predictive analytics km chicago
DESCRIPTION
http://kmchicago.blogspot.com/2010/05/may-11-meeting-using-predictive_06.htmlTRANSCRIPT
© 2010 IBM Corporation
Business Analytics
Predictive Analytics:Overview
1
Jing ShyrChief Statistician, SPSS Predictive Analytics Product Development
© 2010 IBM Corporation
Business Analytics
2
Introducing SPSS, an IBM Company
A leading provider of predictive analytic software, services and solutions
– Software – data collection, text and data mining, advanced statistical
analysis and deployment technologies– Services – implementation, training, consulting, and customization – Solutions – combine software and services to deliver high-value line-
of-business solutions; used for optimizing marketing campaigns, call center effectiveness, identification of fraudulent activity and more
40 years of experience and a broad customer base– 250,000 customers: 100 countries, 50 states, 100% of top universities
Enables decision makers to predict future events and proactively act upon that insight to drive better business outcomes
© 2010 IBM Corporation
Business Analytics
–The world becomes smarter…
Copyright © DreamWorks SKG. 2002.
Vision
3
© 2010 IBM Corporation
Business Analytics
The Predictive Analytics Process
Decision Optimization
Recommend
the most
appropriate
action to take
People Data
& Enterprise Data Sources
Store new data
on customers,
events, etc. for
continuous
improvement
Predictive Analytics
Analyze data to
provide insight and
predict the future
Understand
Predict
ActImprove customer retention
Grow share of wallet
Minimize risk
Increase customer satisfaction
Enhance market share
Prospects
Customers Constituents
Employees
Students Patients
© 2010 IBM Corporation
Business Analytics
C
o
p
y
r
i
g
h
t
2
0
0
3
-
5
,
S
P
S
S
I
n
c
.
5
Close the “Execution Gap”
Close the “Knowledge Gap”
Predictive analytics …
Derives maximum value from its data assets
Understands its business by gaining deep insight
Leverages advanced analytics to predict outcomes
Turns this knowledge into action
Optimize decision making across all operations
© 2010 IBM Corporation
Business Analytics
C
o
p
y
r
i
g
h
t
2
0
0
3
-
5
,
S
P
S
S
I
n
c
.
7
How PA relates to statistics and data mining?
PA uses statistics and data mining to Understand and Predict
Both of them examine and prepare data, apply or try different algorithms for
better prediction
• Stat: Regression, ANOVA, MANOVA, Logistic regression, Discriminant,
Factor, K-mean Cluster, Hierarchical Cluster, generalized linear model, Arima,
…
• Data mining: Neural Network including MLP (Multi-Layer Perceptron), RBF
(Radial Basis function), Kohonen, Bayesian network, Naïve Bayes,
Association, sequence …
• Other: Support vector machine (SVM), Decision Tree, projection pursuit
regression (PPR), nonnegative factorization…
© 2010 IBM Corporation
Business Analytics
Problems lead to decisionsPredictive Analytics Driving Decisions
Customs & Border Protection
– Problem: I can’t search every car that crosses the border.
– Decision: Which car should I search?
Infinity
– Problem: I can’t investigate every claim for fraud.
– Decision: Should I investigate this claim?
Cablecom
– Problem: I can’t save every customer.
– Decision: Is it worth trying to save this customer?
8
© 2010 IBM Corporation
Business Analytics
Healthcare & Insurance Claims Management
What if you could predict fraud before it
happened?
What if you could recommend preventative
care to those who most need it?
What if you could process low risk claims
faster and with less headache?
What if you could plot the expected course
of treatments for veterans?
What if you could react differently in times
of crisis?
© 2010 IBM Corporation
Business Analytics
Two special data sources
Text : unstructured data
Capture customer issues/measure preferences
expressed in survey text, call center notes, and Web
data
Social Network dataCall Detail Record (CDR): A CDR contains all the details
pertaining to a call such as the time, duration, origin, destination, etc.E-mail, Facebook, …
1
0
© 2010 IBM Corporation
Business Analytics
Text Analytics Overview
What does Text Analytics deliver?–Breadth: Take into account qualitative input from all sources–Clarity: Understand related facts, opinions, and what to do about it–Speed: Rapid understanding of qualitative feedback
What does Text Analytics do for people?–Extracts and classifies unstructured data in multiple languages–Discovers patterns in events and opinions and categorizes them–Models customer behavior based on qualitative insights
How does Text Analytics do it?–Natural Language Processing–Sentiment and Event Analysis
© 2010 IBM Corporation
Business Analytics
Horizontal Solution Architecture: Text Analytics
DataAccess
ConceptExtraction
ConceptClassification
RecordScoring
Category Deployment
FileSystem
RDBS
SocialMedia
RDBS
AnalyticalApplication or Tool
Data is
accessed
for
analysis.
Sometimes
data might
be
translated
after being
imported.
Using the
classification
definitions,
records or
documents
are scored.
Using either
manual or
automated
means,
concepts are
classified.
Classification
can be done
on a per
record basis
or on a
concept
basis.
Data is
indexed,
tokenized,
normalized.
Concepts
and text
link
patterns
are
generated.
Categorized
records or
documents
are ready for
further
analysis or
for various
reporting
options.
12
Sentiment Analysis
Dashboard or Presentation Tool
Opinions
(positive
and
negative)
are
associated
with
persons,
places, and
things.
Dashboard or Presentation Tool
© 2010 IBM Corporation
Business Analytics
1
3
IBM SPSS Text Analytics
Uses natural language processing
heuristic rules and statistical
techniques to reveal conceptual
meaning in text
Extracts concepts from text and
categorizes them
Makes unstructured qualitative data
more quantifiable, enabling the
discovery of key insights from
sources such as survey responses,
documents, emails, call center notes,
web pages, blogs, forums and more
Brings repeatability to ongoing decision making
© 2010 IBM Corporation
Business Analytics
Trust Network Described as A Circle Graph
© 2010 IBM Corporation
Business Analytics
Traditional Applications
SNA applies to a wide range of business problems, including:
Knowledge Management and Collaboration. SNAs can help locate expertise, seed new communities of practice, develop cross-functional knowledge-sharing, and improve strategic decision-making across leadership teams.
Team-building. SNAs can contribute to the creation of innovative teams and facilitate post-merger integration. SNAs can reveal, for example, which individuals are most likely to be exposed to new ideas.
Human Resources. SNAs can identify and monitor the effects of workforce diversity, on-boarding and retention, and leadership development. For instance, an SNA can reveal whether or not mentors are creating relationships between mentees and other employees.
Sales and Marketing. SNAs can help track the adoption of new products, technologies, and ideas. They can also suggest communication strategies.
Strategy. SNAs can support industry ecosystem analysis as well as partnerships and alliances. They can pinpoint which firms are linked to critical industry players and which are not.
© 2010 IBM Corporation
Business Analytics
What are the results SNA produce?
Identify groups (communities)
Identify leaders or influencers
Execute viral marketing strategies
Identify product Up-selling and Cross-selling opportunities
Manage contagious churn
Identify subscriber acquisition and retention opportunities
© 2010 IBM Corporation
Business Analytics
1
7
Act: IBM SPSS Decision Management
Framework for domain
specific applications that
combine Models, Rules, and
Optimization to solve
business problems
Extends predictive insights to
the business user at the point
of decision– E.g. Should a claim be ‘fast
tracked’ or evaluated more
closely based on a calculated
risk score?
Automating high volume, high value decisions
© 2010 IBM Corporation
Business Analytics
1. Connect to Data
2. Define Global SelectionsIdentify who or what is to be included as well explicitly excluded from the decision making process
3. Define Desired Outcomes
Define the set of potential decisions that can be made (what campaigns are available, which types
of investigation can be performed etc)
4. Define Operational Decisions with Rules & Models
Define and use rules and/or predictive models that dictate or help decide on the appropriate
outcomes
5. Optimize OutcomesSpecify how the rules and models should be combined to make the most optimal decision
6. Deploy the solutionin batch or for real time decisioning
7. ReportMonitor the decisions that have been deployed through reporting
Best practices approach to decision making
based on our experience in the marketplace
1
8
Repeatable Approach : 7 Steps to Analytical Decision Making
© 2010 IBM Corporation
Business Analytics
Configurability Configurable in the field to new business problems
Enable services / partners to deploy decisioning services to a wide range of business
problems
Terminology is configurable to different applications
– Customer Interactions, Claims, Risk, Churn, Underwriting, Claims, Subrogation etc.
Configurable around the 7 steps
– Which steps are required?
– Various options for working with and combining rules / models and for optimizing
the decision returned
1
9
© 2010 IBM Corporation
Business Analytics
Demo Business Problem – Claims Management
A large insurance company
wants to manage claims
more effectively
–Reduce the time needed to
process a typical claim.
–Reduce the amount paid to
fraudulent claims
The Claims Management
Application processes
incoming claims in real time,
and recommends the best
action for each claim
© 2010 IBM Corporation
Business Analytics
Step 1 & 2: Define Decision Scope…(Sample Illustration: Insurance)
The decision process begins with leveraging enterprise data
and identifying the focus of the operational decision.
excluded.
The Insurance Company elects to exclude data related
to natural phenomenon's.
Application: “I don’t want to worry about Claims associated with Katrina”
© 2010 IBM Corporation
Business Analytics
Step 3: Defining Desired Outcomes…
(Sample Illustration: Insurance)
Typically with all decisions there is a finite set of desired
outcomes that can be achieved.
The Insurance Company identifies three possible
outcomes to the decision.
Application: “There are three things we could do: Fast track, Standard process, Investigate”
This structure can be
multidimensional
© 2010 IBM Corporation
Business Analytics
Steps 4: Define Operational Decisions…(Sample Illustration: Insurance)
Both are critical to optimize outcomes!
Business people define
rules that embody their
priorities and experiences.
Business People
leverage existing
predictive models – or
create new ones, to
support the business
problem.
Application:
“I know that claims for active servicemen go through a serious evaluation before
submittal, so even if the profile is high risk, we can still process it.”
© 2010 IBM Corporation
Business Analytics
Step 5: Optimize Outcomes using Matrix…(Sample Illustration: Insurance)
The decision outcome is optimized and balanced between the
predictive components that provide real time insight and the
rules that govern the policy and practices of the company.
Business people run multiple simulations and identify the best approach
Application: “I wonder what would happen if we evaluated all the claims for fraud?
Hmmm the allocation would overwhelm the department”
© 2010 IBM Corporation
Business Analytics
Step 5: Optimize Outcomes using Formula Approach
The decision outcome can also be
determined by configuring
formulas which will automatically
determine the right action as
projected by rules and models
It’s all controlled by the business
Application: “Recommend preventative care if the risk profile is high”
© 2010 IBM Corporation
Business Analytics
Step 6: Deploy Decision Pattern to Enterprise
Single button deploy alerts IT
that it’s time to move the
solution into production–Point of Interaction Systems can
drive best practices for every real
time decision.
–Automation service can update data
records to reflect operational policy
decisions
–Model Management capabilities
allow ongoing monitoring /
improvement of the models in
production
© 2010 IBM Corporation
Business Analytics
The Report tab allows you to monitor the
status of deployed applications
Application: “How did our new policy impact total claim costs?”
The business can check up on results, and
adjust how things are handled – starting the
process over……..
Step 7: Report on outcomes – and Learn!
© 2010 IBM Corporation
Business Analytics
Summary: Enabling the Business UserOptimizing Operational Decisions for Better Results
Web based business user
interface configurable in the
field to new business problems
Built on Convergence!• Data Mining
• Business Intelligence
• Business Rules
• Event Processing
• Data Management
2
8
© 2010 IBM Corporation
Business Analytics
© Copyright IBM Corporation 2010 All rights reserved. The information contained in these materials is provided for informational purposes only, and is provided AS IS without warranty of any kind, express or implied. IBM shall not be responsible for any damages arising out of the use of, or otherwise related to, these materials. Nothing contained in these materials is intended to, nor shall have the effect of, creating any warranties or representations from IBM or its suppliers or licensors, or altering the terms and conditions of the applicable license agreement governing the use of IBM software. References in these materials to IBM products, programs, or services do not imply that they will be available in all countries in which IBM operates. Product release dates and/or capabilities referenced in these materials may change at any time at IBM’s sole discretion based on market opportunities or other factors, and are not intended to be a commitment to future product or feature availability in any way. IBM, the IBM logo, Cognos, the Cognos logo, and other IBM products and services are trademarks of the International Business Machines Corporation, in the United States, other countries or both. Other company, product, or service names may be trademarks or service marks of others.