Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Contents
NOTE: This deck has been designed to provide the elements needed to construct a tailored
presentation to a prospect/customers requirements.
It is not intended to be used in full.
WHAT IS PREDICTIVE
ANALYTICS?
3
Engaging
Addressing the
Analytical Needs of
the Business User
Breakthrough
Analytics for All Data
Exploiting Value
From the Relevant
New Mix of Data
Pervasive
Making Decisions
At Point of Impact
250%
2
1. Divide & Conquer: Using Predictive Analytics to Segment, Target & Optimize Marketing (pg. 1), Aberdeen, February 2012.
2
Source: IDC, The Business Value of Predictive Analytics, June 2011
63%
of organizations realize
a positive return on their
analytic investments
within one year
46%
Predictive
Maintenance and
Quality (PMQ)
Predictive Customer
Intelligence (PCI)
Counter Fraud
Management
(CFM)
Custom Applications
Data
Collection
Statistics
Modeler and
ADM
Analytic
Server
Decision
Optimization
Watson
Analytics
SOLUTION BUYER
SELF-SUFFICIENT BUILDER
SELF-SERVICE USER
Puts predictive
power into the
hands of a
business analyst
Provides the
sophistication needed
by an expert
Includes a range of
advanced data
manipulation and
analytical algorithms
Flexible deployment
options
IBM Provides
Business Process
as a Service
Software
as a Service
Customer/User is looking to
move to cloud and wants IBM to
manage infrastructure and
applications it uses
Platform
as a Service
Infrastructure
as a Service
On-premises
Customer/User is looking to
move to cloud and wants IBM to
manage infrastructure
Hybrid Offering
On-Premises or On-Cloud
Rules
Applications Data
Predictive
Analytics
Time Series
Geotemporal
and Geospatial
Relational
Social Networks
Simulation and
Optimization
Scoring
Business
Rules
Predictive
Analytics
Modeler
Server
Analytical
Decision
Management
Statistics
Server
Real Time
Scoring
Client
Software
Optional
Hosted environment
with software,
security and
infrastructure
managed by IBM
Subscription pricing
with flexible terms
Algorithms
Usage
Classification
(Or Prediction)
Autoclassifiers, Decision
Trees, Logistic, Support
Vector Machines, Time Series
Segmentation
Autoclusters, K-Means,
Anomaly Detection
Association
Geospatial
Space-Time Boxes
Automated
Autoclassified, Autonumeric,
Time Series, Clustering
Simulation
Monte Carlo
Specialized
In-database
Open Source
CLEF
Unlimited
MODELING ALGORITHMS
Techniques included
Decision Trees
Bayesian Networks
Neural Networks
Decision List
Statistical Models
Time Series
Self Learning Response Models
Support Vector Models
Nearest Neighbor Models
Segmentation
Techniques included
Kohonen
K-Means
TwoStep
Association
Automated Modeling
Geospatial analytics
Apriori/Association
TCM (Temporal Causal Modeling)
STP (Spatio Temporal Prediction)
TwoStep Cluster
Bill Smith
123 Main Street
(800) 555-1212
SSN: 444-33-2222
DOB: 8/7/84
Applicant: Today
William R Smith
123 S Main Avenue
(100) 111-1234
DL: 90909091
DOB: 7/8/84
Arrested: Feb 2013
Call Center
Complaint
Not Actionable
Influential
@Twitter
Terminated
Employee
Context Accumulation: The incremental process of integrating new observations with previous
observations
Entity Analytics
Information In Context
Observation
Space
Consumption
Name
Beth102
L. Johns
Entity
-Parker
BL Johns
Addr1
123 Main Street
777 Park Road
City
New York
State
NY
Phone
2127331234
DOB
6/21/1954
Income
$8,000
Credit Debt
$5,359
Other Debt
$2,009
Debt to Income
92.1
Prev Default?
True
Pending Loan
False
Full
Liz Johns
Addr1
33 Red Dr
City Entity 343
Mamaronec
k
State
NY
Postal
10354
Phone
212-7331234
914-6982234
Income
$9,000
Credit Debt
$6,000
Other Debt
$3,000
Debt to Income 100
Prev Default?
True
Pending Loan
False
Full
Entity 642
Elizabeth Lisa
Johns
Addr1
33 Reed Dr
City
White Plains
State
NY
Postal
10354
Phone
914-698-2234
Income
$31,000
DOB
6/21/1954
Credit Debt
$1,362
Other Debt
$4,001
Debt to Income 17.3
Prev Default?
False
Pending Loan
True
Resolved Entity
Name
Elizabeth Lisa
Johns
Liz Johns
Beth L JohnsParker
BL Johns
Addr1
123 Main Street
777 Park Road
33 Red Dr
33 Reed Dr
City
New York,
White Plains,
Mamaroneck
State
NY
Postal
11732, 10354
Phone
212-733-1234
914-698-2234
DOB
6/21/1954
Defaults
Yes
Income
$48,000
Credit Debt
$12,722
Other Debt
$9,009
Debt to Income 113.5
Prev Default?
True
Pending Loan
True
Persist Searches
Optionally add new
streaming records
to the repository
Anonymization
Anonymize features in the repository
to address privacy concerns
Text Analytics
STAfS
Yes
No (needs to be
Text format
Short/long responses,
documents and folders of
documents
Amount of text
Small to large
(can use Modeler Server)
Smaller sets
generally up to
10,000 records
Sentiment analysis
Yes
Yes
Yes
No
Yes
No
Yes
No
No
Churn Prediction
Marketing
Govern
Disciplined approach instills confidence
Managed changes
Audit ready
ANALYTICS AT SCALE
Velocity
Data at Rest
Data in Motion
Terabytes to
Exabytes of existing
data to process
Streaming data,
milliseconds to
seconds to respond
Variety
Data in
Many Forms
Structured,
unstructured, text,
multimedia
Veracity
Data in Doubt
Uncertainty due to
data inconsistency
& incompleteness,
ambiguities, latency,
deception, model
approximations
SPSS Modeler
SPSS Modeler
Server SPSS
Analytic Server
Database/Hadoop
Data Preparation
and Model
Building/
Scoring
Pushback
Server
Resources
Used for
Analysis
Data At
Rest
Data In
Motion
SQL pushback
In-database mining
IBM InfoSphere
Warehouse
IBM PureData System
for Analytics (Netezza)
Oracle
SQL Server
In-database adapters
IBM PureData System
for Analytics (Netezza)
IBM DB2 for z/OS
Teradata
Parallel processing
IBM InfoSphere BigInsights
Other Hadoop distributions
Data
Database Resources
Used for SQL
Pushback, In-DB
Processing and
Map/Reduce
Processing
Server Resources
Used for Analysis
Predictive analytics for big data IBM SPSS predictive analytics and
IBM PureData System for Analytics
Visual, easy-to-use interface
Faster time to solution and understanding
Accessible analytics for a business user,
sophistication and power for the data scientist
< 4 Seconds
100M Customers
1 Model
10 Predictors
< 10 seconds
100M customers
20 models
20 predictors
Predictive analytics for big data IBM SPSS predictive analytics and
IBM PureData System for Analytics
ICU
Monitoring
Cyber
Security
Powerful
Analytics
Algo
Trading
Millions of
Events per
Second
Government/
Law Enforcement
Telco Churn
Prediction
Smart
Grid
Microsecond
Latency
Traditional/Non-traditional
Data Sources
No database extensions
required
Requires database
extensions to be installed
Performance/reliability
harder to predict
Performance/reliability
easier to predict
Di
Predictive
Maintenance and
Quality (PMQ)
Predictive Customer
Intelligence (PCI)
Counter Fraud
Management
(CFM)
Custom Applications
Data
Collection
Statistics
Modeler and
ADM
Analytic
Server
Decision
Optimization
Watson
Analytics
Cognos BI
Cognos Package
Consume Analytics
Report / Dashboard
Data Preparation
SPSS Modeler
SPSS Modeler
Create Report /
Dashboard
Export Data
Author Report
Cognos Package
TM1 Integration
SQL / UDF
Modeler Client
Relational Database
IBM SPSS
Analytic Server
Modeler Server
Hadoop Job
Analytic Catalyst
Tablet Client
Analytic Catalyst
Browser Client
Analytics
Modeler supports integration with IBM Netezza, providing the ability to run data
mining algorithms to be directly in the IBM Netezza environment from the Modeler
user interface.
The following algorithms from Netezza Analytics are supported within Modeler
Bayes Net
Decision Trees
Divisive Clustering
Generalized Linear
K-Means
KNN
Linear Regression
Naive Bayes
PCA
Regression Tree
Time Series
2 Step cluster
Environment
Monitoring
ICU
Monitoring
Powerful
Analytics
Algo
Trading
Cyber
Security
Millions of
Events per
Second
Government /
Law Enforcement
Telco Churn
Prediction
Smart
Grid
Microsecond
Latency
Traditional / Non-traditional
Data Sources
Extend
Basic predictive
capabilities for the
business. Go beyond
spreadsheets and
quickly explore data in
the context a business
can understand
Collaborate
Meaningful analytics that
a novice begins and an
expert builds upon.
Bring business led
insights to decision
makers and advanced
analysts.
Transition
From discovery to rich
story telling capabilities,
embed predictive findings
into decision management
models for optimal
business efficiencies.
Decision Optimization
Cost vs.carbon
emission
Prescriptive
Consumer
Goods
Retail
Manufacturing
Statistics Approach
Modeling Approach
Allows Statistics models, transformations, output and syntax within the Modeler
GUI
Statistics dialog boxes for consistency
Uses Statistics in the background to run analysis from the Modeler interface
One way integration i.e. Statistics can be used within Modeler, but not vice versa
Requires a Statistics license for the procedures