Feature Engineering Studio

Download Report

Transcript Feature Engineering Studio

Feature Engineering Studio
September 9, 2013
Welcome to
Problem Proposal Day
• Rules for Presenters
• Rules for the Rest of the Class
Rules for Presenters
• Talk for 3 minutes on:
– Data set
– What variable will you predict?
– What kind of variables will you use to predict it?
– Why is this worth doing?
• Remember to send me your slides (if any)
Rules for Audience
• After the presentation
– Ask quick questions
– Give quick suggestions
• Everyone
– Is the problem genuinely important? (usable or
– Is there a good measure of ground truth?
• Only if you know what you’re talking about
– Is there rich enough data to distill meaningful
– Is there enough data to be able to take advantage
of data mining?
Rules for Audience
• Be polite!
• No interrupting
• No rambling
• No being mean
First Step
• Get into the right collaborative spirit
• You are officially encouraged
(though not required)
to sing along
• http://www.youtube.com/watch?v=pd_5-2kCzfs
– 0:25
• Alphabetical Order Based on Last Name
– Tie-Breaker: First Name
For next week
• Think about how to improve your problem
• Rewrite your problem proposal based on the
feedback you got today
• Then email it to me for further feedback and a
“thumbs-up” before the next class
Assignment 2
• Data Familiarization
“Mucking Around”
Get your data set
Open it in Excel
Look at your ground truth label (if you have one)
Look at other key variables
• What does each variable mean semantically?
• If numerical, what are its max, min, average, stdev? Create
histograms of key variables.
• If categorical, what is the distribution of each value?
Assignment 2
• Data Familiarization
“Mucking Around”
• Write a brief report for me
• You don’t need to prepare a presentation
• But be ready to discuss what you learn about
your data
What if you don’t have data yet?
1. Get your data
2. If you can’t get your data before class, email
me at least 48 hours before class and I’ll send
you a practice data set
How to compute in Excel
• If numerical, what are its max, min, average,
• If categorical, what is the distribution of each
• Using Class2Data
How to do a histogram in Excel
• Using Class2Data
Next Class
• 9/23 Feature distillation in Excel (Asgn.2 due)
– Do the assignment
– Read the readings
Upcoming Classes
• 9/23 Feature distillation in Excel (Asgn.2 due)
• 9/25 Special session on prediction models
– Come to this if you don’t know why student-level crossvalidation is important, or if you don’t know what J48 is
• 9/30 Advanced feature distillation in Excel (Asgn. 3
• 10/2 Special session on RapidMiner
– Come to this if you’ve never built a classifier or regressor
in RapidMiner (or a similar tool)
– Statistical significance tests using linear regression don’t