Better Analysis, Deeper Insights: A Public Sector Primer on Data Mining

1
Better Analysis, Deeper Insights  A Public
Sector Primer on Data Mining

• Jennifer Galvan
• Manager, Sales Engineering

2
Commonly Asked Questions

• Will I be able to get copies of the slides after
  the event?
• Is this webinar being taped or can I view it
  after the fact?

Yes
Yes
www.spss.com/events
3
Agenda

• Predictive Analytics
• Data Mining and Text Mining
• Government applications
• Data Mining Methodology
• Clementine
• Getting Started

4
SPSS Fundamentals

• Founded in 1968
• 30 year heritage as an innovator in analytics
  technologies
• IPO in 1993 (NASDAQ SPSS)
• Operations in more than 60 countries
• 16 organizations acquired since 1993
• Leadership
• Market leader in predictive analytics
• Recognized as a leader by Forbes, BusinessWeek,
  Intelligent Enterprise, InfoWorld, CRM Magazine,
  and others
• Vital statistics
• 261.5 million in revenue in fiscal year 2006,
  11 increase
• 1,200 employees around the world
• 250,000 customers in business, academia, and
  government

042806
5
What is Predictive Analytics?
Predictive analysis helps connect data to
effective action by drawing reliable conclusions
about current conditionsand future
events. Gareth Herschel, Research Director,
Gartner Group
6
Predictive Analytics can be leveraged to enhance
decisions within the business
Predict
Analyze data to provide insight and predict the
future
Recommend the mostappropriate actionto take
Predictive Analytics
Capture
Act
?Improve customer retention ?Grow share of
wallet ?Minimize risk ?Increase customer
satisfaction ? Enhance market share
Customers
Constituents
Store new data on customers, events, etc. for
continuous improvement
Prospects
Employees
Customer View
Students
Patients
Decision Optimization
People Data Enterprise Data Sources
7
Data Mining Defined

• Data driven approach to problem solving
• Focused on Business Objectives
• Leverages organizational data
• Uncovers patterns using predictive analytics
• Uses results to help improve business decision
  making and organizational performance

8
What is Data Mining?

• Discovering meaningful patterns in your data

9
What is Data Mining?
As the data grows

the relationships become more complicated.
10
Data Mining and Statistics

• Statistical Analysis
• Tests for statistical correctness of models
• Are statistical assumptions of models correct?
• Hypothesis testing
• Is the relationship significant?
• Tends to rely on sampling
• Techniques are not optimized for large amounts of
  data
• Requires strong statistical skills

• Data Mining
• Less interested in the mechanics of the technique
• If it works and makes some sense, lets use it
• No assumption required
• Can find patterns in very large amounts of data
• Requires understanding of data and business
  problem
• Focus on Deploying Results

11
Government and Citizen Understanding

• What are the voters saying about me?
• I know that people in this neighborhood are
  unhappy but why?
• We always talk about crime and healthcare, but
  are these the main issues on peoples minds?
• Young people dont vote. How can we get them
  engaged?
• What is the real problem with our schools?

12
Unstructured Data Holds the Keys but do you use
it effectively?

• Most just ignore qualitative customer input
  provided in surveys, emails, phone calls, and
  other sources its not even captured.
• Many store important business/customer text
  information but have no way to leverage it
• A few start mining text but are unable to match
  results on business data

13
State of Texas

• What data mining has done for

Challenge Improve tax audit selection process
Recovered 400 million in unpaid taxes
14
Tax Gap

• Tax administrators need analytics to cope with
  an increasing tax gap with limited resources
• Technology adoption is on the rise
• Non-filer discovery TX, MA, CT, NM, IA, OK
• Audit selection IRS, TX, NY, (SC)
• Collections CA, VA, KS, MO
• Many benefits-based projects (self-funded)
• The scale of potential payback is very large
• TX (discovery) 400M over 5 years
• VA (collections) Break even after 2 months
• MA (discovery) 88M/week

15
Centers for Medicare Medicaid Services (CMS)
-

• What data mining has done for

Challenge develop a program that will isolate
factors that lead to incorrect Medicare payments
Reduced payment errors by 50.
16
Common Applications in Public Sector

• Voter concerns
• Law Enforcement
• Fraud, Waste and Abuse
• Education

17
Put yourself in this position

• You are totally new to data mining
• and none of your colleagues have done it before
  either
• but your agency has decided its a Good Thing
• and they want you to lead the development of a
  predictive analytics approach to decision making.
• Where do you start?

18
The Methodology - CRISP-DM

• 6 Phases
• Business Understanding
• Data Understanding
• Data Preparation
• Modeling
• Evaluation
• Deployment
• Reflects iterative nature of data mining

19
Getting started with Data Mining

• Follow CRISP-DM
• Begin with the end in mind
• Limit the scope of your initial project
• Define an executable data mining strategy
• Line up the right people
• Line up the right data

20
Clementine Market Position

• Clementine is the leading data mining workbench
  because it is
• Easy to use
• Comprehensive
• Supports the entire data mining process
• Provides outstanding performance scalability
• Therefore delivers
• High productivity
• Quick time-to-solution
• High ROI

21
Clementine Ease of Use and Comprehensive
Facilities

• Clementines visual approach makes it easy to
  integrate a comprehensive range of facilities
  while remaining problem oriented
• Suitable for the Business Analyst as well as the
  Technical Expert

22
Visual Data ManipulationDedicated nodes for
extensive list of techniques

• Record Operations

• Field Operations

23
Automation Numeric Predictor

• Automated modeling operations create evaluate
  many different models in one step
• New Numeric Predictor node means automated
  modeling for numeric outcomes

24
Binary Classifier

• Binary Classifier is automated modeling for
  yes/no outcomes

25
Graphboard

• Graphboard provides a range of visualizations,
  both traditional and advanced
• Two modes
• Basic mode acts like a wizard select data
  and get offered a range of appropriate graphs
  options
• Detail mode traditional pattern select
  graph, select data, select options

26
Custom Tables

• Create complex reports / nested tables
• Designed using pivot table or drag and drop
  style user interface

27
Getting Started

• Data Mining Jump Start
• Reduce time to implement your first data mining
  project
• Combination of training and coaching
• 5 day on-site

28
Question and Answer
29
For More Information

• In case you missed it recorded version and
  slides available at www.spss.com/events
• Visit www.spss.com/clementine to learn more about
  the platform
• Call us at 1-800-543-2185 or sales_at_spss.com
• Please fill out the post event survey