Next Gen Data Mining: A Personal Viewpoint

1
Next Gen Data MiningA Personal Viewpoint

• Rakesh Agrawal

2
Thesis

• Data mining should aspire to become the
  technology that allows entities to derive maximal
  economic value from data while respecting the
  sovereignty of data

3
Quest What did we get right (if any)
4
Abstraction of data mining operations from
applications
5
Sins we committed
6
Too timid in our ambition for data mining
7
Didn't emphasize composability and integration
8
Didn't desgin-in sovereignty of data
9
Two contradictory views of Data Mining
10
Not very valuable in practice
11
Too powerful Ban it
12
What gives?
13
The way world sees it
14
Collect Organize Data
Province of Data Management
15
Extract value from Data
Province of Data Mining
Model Building Application Complex
Querying Search
16
What should we do?
17
Accept the mandate
18
Approach
19
Start with High-Value Applications
Distill Abstractions
Don't fall in the Quest traps
Don't limit to structured data Consider rich
operations Keep sight of composability
integration Design in data sovereignty
20
An Application
21
A grand challenge application

• Here is my data
• This is what I already know (priors)
• This is how I assign value
• Find for me that is valuable and
• too frequent
• too infrequent
• new and has changed since last interaction
• Allow me to intervene and guide the process
• Do it across sovereign entities

22
Thesis

• Data mining should aspire to become the
  technology that allows entities to derive maximal
  economic value from data while respecting the
  sovereignty of data

23
Your tomatoes now
24

• Next Generation Data mining should aspire to
  become the technology that allows entities to
  derive maximal economic value from data while
  respecting the sovereignty of data.
• Here is a grand challenge problem to drive the
  future development of data mining
• Here is my data
• This is what I already know (priors)
• This is how I assign value
• Find for me that is valuable and
• too frequent
• too infrequent
• new and has changed since last interaction
• Allow me to intervene and guide the process
• Do it across sovereign entities