Table of Contents
Data Mining Business Hunching, not Data Crunching
The story board
The context of data mining
What is Data Mining?
Data holds Knowledge
What can we learn from data?
What is data mining?
An example Winterthur Insurance, Spain
Winterthur Insurance Application
Winterthur Approach
Winterthur Application
Winterthur Application
Winterthur “gains” curve
Data mining users - Finance
Manufacturing Utilities, Defence
Telecommunications
Pharma, Chemical, Retail
Government, Media, Consulting, IT
Halfords - Predicting Sales
Reuters Forex Data
Typical Applications
Typical Applications
Typical Applications
Typical Applications
The chasm of representation
The data is not the business
Fundamental DM Assumptions - 1
Many fuzzy things: hard to model
Hunching with the data
Fundamental DM Assumptions - 2
Business knowledge is crucial
Data Mining is easy
But data mining is complex
Many suppliers, hype, confusion
Data Mining is
Hydraulic data mining
Bikers and seafood?
Hydraulic data mining
Good Data Mining is:
Understand the Business Problem First
DM Process in a multi-level IS
Decision models
Models deployed in an IS Decision models (“agents”) in action
PPT Slide
Mysterious disappearing terabytes
PPT Slide
Data Mining Technology
Terminology
Data Exploration
Visualisation
Machine Learning
The Technology
Machine Learning
Neural Networks
Neural Networks
Problems
Rule Induction
Understanding Rule Trees
Rules & Nets
Clustering - Kohonen Networks
Some techniques are “automatic”
Some build executable models
Technology limitations
BBC TV Audience Prediction
Results Assessment
Danger of believing “explanations”
Rationale for ClementineTM
Bridging the chasm of action
Clementine: beware of imitations!
Now the real Clementine
“Hunching, not crunching”
Essay Question
The End
|