Top ten algorithms in data mining (2007) [pdf]

Re:Top ten algorithms in data mining (2007) [pdf] by _delirium on Tuesday December 18, 2012 @01:33PM (#4938787)

They were one of the first widespread ensemble methods (based on bagged decision trees), and share with other bagging/boosting methods fairly good empirical performance on a wide range of problems.

[ Parent | on HN ]

Re:Top ten algorithms in data mining (2007) [pdf] by tocomment on Tuesday December 18, 2012 @01:39PM (#4938820)

Are they easy to interpret? Is it ok if I have a lot of columns in my data?

[ Parent | on HN ]

Re:Top ten algorithms in data mining (2007) [pdf] by textminer on Tuesday December 18, 2012 @01:52PM (#4938882)

You will find no understanding in the grown trees-- by definition, they're each seeing a different angle of the data, and each node a different subset of available features.

You can, however, calculate importance scores for the features used. Brennan's original paper gives a good algorithm for doing this (in short: for each tree, permuting the data along some feature for an out of bag sample and seeing how much worse it does.)

[ Parent | on HN ]

Re:Top ten algorithms in data mining (2007) [pdf] by textminer on Tuesday December 18, 2012 @02:03PM (#4938964)

(Pardon me/blame autocorrect: Breiman.)

[ Parent | on HN ]

Re:Top ten algorithms in data mining (2007) [pdf] by glogla on Tuesday December 18, 2012 @01:47PM (#4938846)

Depend on the number of trees, and their depth. Even with 20 trees, with effort, it could be written on paper as rules, for example. And it helps if most of the trees are depth 3 or something.

But then, some models are pretty much impossible to understand or interpret.

[ Parent | on HN ]

Re:Top ten algorithms in data mining (2007) [pdf] by _delirium on Tuesday December 18, 2012 @02:33PM (#4939127)

They handle large-dimensional spaces fairly well, which is one of their strengths. They are not all that interpretable, though.

[ Parent | on HN ]