CONFIDENTIAL
Advanced Analytics
Business Intelligence with
Data Mining
Data Mining
Whats important
Association/Binning
Clustering
Classification
Segmentation
What to expect
What-if
Estimation
Curve Fitting
Fill in Sparse Matrix
Prediction
Probability
Quantitative
Methodology
Statistical Analyst Business Modeling
Collected Sample
Data
Store
Predictive Metrics & Segments
DBA
business interpretation
Marts
Warehouse
Optimize data marts
Methodology - EDMDAPA
Extract
Integrate disparate data systems
Build holistic business view
Group and organize large sets of categorize
Discretize/Classify
Grouping and Segmentation
Simplify large flat dimensions
Model
Create predictive estimation functions
Deploy
Build/score data marts, cubes with predictive probability and quantitative metrics
and simplified dimensional categories
Analyze, Visualize, Scorecard
Identify KPI's, Identify business problems
Plan
Predict(Forecast)/Test(What-If)
Apply performance rules on KPIs
Act
Campaigns, personalization, optimization
Extract
DecisionStream unites information from disparate data
sources for sampling the enterprise
80% of the work involved in analytics is collecting,
cleansing, and preparing data
Classification with Scenario
Segment and
Classify
combinations of
stores, regions,
divisions, customers
or products
Benchmark against
last month!
Path of success
Model with 4Thought
Avoids over-fitting
Works well with
Noisy
Co-linear
Not much or sparse data
Factor Analysis
What-if
Filling in the sparse matrix e.g. #1
Revenue estimation:
Dimensional intersect:
Red shoes, southwest, women, springtime:
$50,000
Black shoes, northeast, men, summer:
$38,000
Black shoes, southwest, women, summer:
$43,000
Black shoes, northeast, men, springtime:
????
Once a model is build against historical data, the resultant
function can productively fill in the question marks
Filling in the sparse matrix e.g. #2
Insurance cost estimation:
Dimensional intersect:
Age 38, southwest, female, non-smoker, married:
$1,800
Age 24, northeast, male, smoker, single:
$2,300
Age 32, southwest, female, smoker, single:
$3,000
Age 28, southwest, men, non-smoker, married:
????
Once a model is build against historical data, the resultant
function can productively fill in the question marks
Deploy with DecisionStream
DecisionStream uses predictive function from
4Thought as UDF for derivation
Deploy data marts, cubes, and metadata
Analyze, Visualize, Scorecard
Plan
Determine Business Goals
and apply
NoticeCast Agents
KPI Business Pack
Exception highlighting with
reports
Forecast with 4Thought
Access forecasted results with
ETL
Keys to Mining
Usefulness
Can the information discovered be
considered knowledge?
Certainty
How viable is the discovered
knowledge
Expressiveness
Can the discovered knowledge be
represented in a meaningful way
Problems for Mining
Missing data
Inconsistent categories
Too much data
Difficult to focus
Not enough data
Nothing meaningful
Too many patterns
Hard to discern knowledge from garbage
Complexity of discoveries
Knowledge is too complex to be used
Unavailable data
The Cognos BI Solution
Integrating touch-points leads to a 360-degree view of your business.
Many scored metrics are loaded via predictive models.
Segmentation is useful for simplifying large flat dimensions.