Return to Video

DT Strengths and Weaknesses - Intro to Machine Learning

  • 0:01 - 0:05
    You did it again, congratulations, your third machine algorithm decision trees,
  • 0:05 - 0:06
    you got all the way to the end.
  • 0:06 - 0:08
    >> We've learned a lot about decision trees.
  • 0:08 - 0:11
    Let's add some of this new found knowledge to our list of things to
  • 0:11 - 0:13
    consider when you're picking a classifier.
  • 0:13 - 0:16
    >> So they're really easy to use and they're beautiful to grow on.
  • 0:16 - 0:18
    They're, they're graphically, in some sense,
  • 0:18 - 0:20
    allow you to interpret the data really well, and
  • 0:20 - 0:24
    you can really understand them much better then say the result of a support vector machine.
  • 0:24 - 0:25
    But they also have limitations.
  • 0:25 - 0:26
    >> That's right.
  • 0:26 - 0:28
    One of the things that's true about decision trees is
  • 0:28 - 0:29
    they're prone to over fitting.
  • 0:29 - 0:32
    Especially if you have data that has lots and lots of features and
  • 0:32 - 0:35
    a complicated decision tree it can over fit the data.
  • 0:35 - 0:38
    So you have to be careful with the parameter tunes that you're picking when you
  • 0:38 - 0:39
    use the decision tree to prevent this from happening.
  • 0:39 - 0:41
    >> Yeah. What comes out could look crazy if your node has,
  • 0:41 - 0:43
    only did single data point.
  • 0:43 - 0:44
    You almost always over-fit.
  • 0:44 - 0:47
    So it is really important for you to measure how well you're doing,
  • 0:47 - 0:50
    then stop the growth of the tree at the appropriate time.
  • 0:50 - 0:52
    >> One of the things that's also really cool about decision trees,
  • 0:52 - 0:55
    though, is that you can build bigger classifiers out of
  • 0:55 - 0:58
    decision trees in something called ensemble methods.
  • 0:58 - 0:59
    In the next lesson,
  • 0:59 - 1:02
    we'll give you a chance to actually explore an algorithm completely on your own.
  • 1:02 - 1:06
    And a couple of the ones that will give you as choices are examples of
  • 1:06 - 1:06
    ensemble methods.
  • 1:06 - 1:08
    So if this sounds interesting to you.
  • 1:08 - 1:11
    Building a classifier out of classifier, then stay tuned in the next lesson.
  • 1:11 - 1:16
    >> Yeah, we are super, duper So stay tuned.
  • 1:16 - 1:17
    >> Let's get started.
Tytuł:
DT Strengths and Weaknesses - Intro to Machine Learning
Opis:

more » « less
Video Language:
English
Team:
Udacity
Projekt:
ud120 - Intro to Machine Learning
Duration:
01:18

English subtitles

Revisions Compare revisions