sas visual statistics...sep 04, 2014 · statistics deployment options distributed commodity hw (*)...
TRANSCRIPT
Copyr i g ht © 2014, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
SAS®
VISUAL STATISTICS
PETER HUGHES
QUEST Q3
SEPTEMBER 2014
CONNECT WITH ME
Copyr i g ht © 2014, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
PREDICTIVE
ANALYTICS FORWARD THINKING
Higher Decision Impact
Monitor & Detect
Current State
Predict & Act
Future/New Opportunities
Copyr i g ht © 2014, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
SAS®
VISUAL
STATISTICS FUELING THE DATA TO DECISION LIFECYCLE
SAS® Visual Statistics
TEXT COMPETITIVE
ADVANTAGE
MANAGE
DATA
EX
PL
OR
E
DA
TA
EXPLORE &
DEVELOP MODELS
DE
PL
OY
&
MO
NIT
OR
SAS® Visual Analytics
Copyr i g ht © 2014, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
SAS®
VISUAL
STATISTICS INTERACTIVE EXPLORATION AND PREDICTIVE MODELING
EXPLORE AND
DISCOVER PREDICT AND
REFINE
COMPARE AND
ASSESS
Copyr i g ht © 2014, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
SAS®
LASR™
ANALYTIC SERVER
“It is an in-memory engine specifically engineered for the
demands of interactive and iterative analytics”
• In-memory = Fast, sub-second responses
• Multi-User = Hundreds of concurrent users
• Stateless = Don’t pre-compute things
• Interactive = Instantly visualize the impact from changing
model parameters
• Deployment = MPP (distributed) or SMP (single machine)
Copyr i g ht © 2014, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
SAS®
VISUAL STATISTICS
ANALYTIC CAPABILITIES HOW IT CAN BE USED?
Classification Predict outcomes such as machine
failure, high-risk patients, etc.
Regression Estimate outcomes such as customer
spend, policy premium, credit limit, etc.
Clustering Segment your data based on self-
similarity to augment your models
Group-By Models by segments/groups (e.g.
location, store, owner, device, etc.).
Copyr i g ht © 2014, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
APPLICATIONS SHOW ME THE MONEY
Predictive Asset
Maintenance Fraud Credit Risk
Customer Segmentation Targeted Acquisition /
Retention / Attrition
Copyr i g ht © 2014, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
SAS®
VISUAL
STATISTICS MANIPULATE DATA
• Access structured and unstructured data
• Data filtering, including outliers
• Join/promote tables, compute columns
• Dynamic Group-By operations
Copyr i g ht © 2014, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
SAS®
VISUAL
STATISTICS EXPLORE DATA
• Discover relationships between variables and augment
model building process
• Derive models directly from correlation matrices, scatter
plots, & box plots
• Visualize results from the modeling process
• Understand individual variable’s level of influence for
all models
Copyr i g ht © 2014, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
SAS®
VISUAL
STATISTICS MODELING TECHNIQUES
• Predictive Techniques
• Linear Regression
• Logistic Regression
• Generalized Linear Model
• Classification Trees
• Descriptive Techniques
• Clustering
• Group-By Processing
• Auto-update
Copyr i g ht © 2014, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
SAS®
VISUAL
STATISTICS RAPID MODEL BUILDING AND REFINEMENT
Copyr i g ht © 2014, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
SAS®
VISUAL
STATISTICS ASSESS AND SCORE
• Model comparison using lift charts, ROC charts,
misclassification tables etc.
• Interactively evaluate lift at different depths of file
• Interactively define event probability cut-off
• Generate SAS code for scoring purposes
Copyr i g ht © 2014, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
SAS®
VISUAL
STATISTICS KEY BENEFITS
Spend more time perfecting your models to reflect
changing conditions and less time waiting for answers
In-memory Analytics provides speed, scale and
concurrency for timely insights
PRECISION
AGILITY
SPEED
Best-in-class data discovery and analytics to derive
precise insights and make targeted decisions
Copyr i g ht © 2014, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
SAS®
VISUAL
STATISTICS DEPLOYMENT OPTIONS
Distributed
Commodity HW(*)
Hadoop HDFS
Cloudera
Hortownworks
Multiple asymmetric
sources
Teradata /
Pivotal /
Oracle
N/A
Asymmetric
Teradata / Pivotal / Oracle
Non-Distributed
Commodity HW(*)
*Virtualization deployment supported with
commodity hardware paths only.
Hardware
Co-located
data store
Asymmetric
source
Copyr i g ht © 2014, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
SAS®
VISUAL
STATISTICS LINEAR REGRESSION
Copyr i g ht © 2014, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
SAS®
VISUAL
STATISTICS LOGISTIC REGRESSION
Copyr i g ht © 2014, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
SAS®
VISUAL
STATISTICS DECISION TREE
Copyr i g ht © 2014, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
SAS®
VISUAL
STATISTICS CLUSTERING
Copyr i g ht © 2014, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
SAS®
VISUAL
STATISTICS MODEL ASSESSMENT
Copyr i g ht © 2014, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
MPP DATASTORE
BLADE ENVIRONMENT
HIGH LEVEL
ARCHITECTURE
DISTRIBUTED DEPLOYMENT ON COMMODITY HARDWARE
(DEDICATED RACK)
IN-MEMORY STORE
SAS® LASR™ ANALYTIC SERVER
SAS® VISUAL ANALYTICS and SAS
® VISUAL STATISTICS
Not part of VS or
VA
Can be separated
TERADATA / PIVOTAL / ORACLE / HADOOP
SAS Embedded Process
WORKSPACE SERVER
MID-TIER
METADATA SERVER
Hadoop HDFS
Cloudera,
Hortonworks
Other RDBMS Nonrelational Click Stream PC Files
WED BASED CLIENT
Copyr i g ht © 2014, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
SAS®
VISUAL
STATISTICS
RESOURCES
EXTERNAL WEB PAGE
Copyr i g ht © 2014, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
VISUAL STATISTICS MORE INFORMATION
SAS Com Visual Statistics
http://www.sas.com/en_us/software/analytics/visual-statistics.html
Attend a FREE Visual Statistics Hands on Workshop
Next one MONDAY 29th September 3pm
Or Just GOOGLE SAS Visual Statistics
Or Youtube SAS Visual Statistics….Lots of information already
Copyr i g ht © 2014, SAS Ins t i tu t e Inc . A l l r ights reser ve d . sas.com
QUESTIONS