YouTube

Got a YouTube account?

New: enable viewer-created translations and captions on your YouTube channel!

English subtitles

← A New Enron Feature - Intro to Machine Learning

Get Embed Code
4 Languages

Showing Revision 3 created 05/25/2016 by Udacity Robot.

  1. Let's start with the example of adding a new feature.
  2. Specifically, I'll be adding some features for the Enron data set and walking
  3. you through how you can figure out whether you're moving in the right direction.
  4. Here's a process I usually follow when coding up a new feature.
  5. The first thing is to start with your human intuition.
  6. What feature do you think might contain some pattern that could
  7. exploit using machine learning?
  8. An example of a human intuition I might have on the Enron data set is
  9. that persons of interest might have particularly strong email
  10. connections between each other.
  11. In other words, they send each other emails at a higher rate than the,
  12. than people in the population at large send emails to persons of interest.
  13. The next thing to do is actually code up the feature.
  14. This is what you'll do next, in a quiz.
  15. After that, visualization will tell you if you're on the right track.
  16. Does your new feature give you discriminating power in the classification
  17. problem we were trying to solve?
  18. And then presumably through this process you've learned something new.
  19. If for no other reason than because you have this new feature available to you.
  20. And very often you'll repeat this process a few times as you zero in on what you
  21. think will be the most helpful new feature for you