Is it important to scale all the features into a common range (normalized) when using random forests (bagging) in classification. Or can random forests handle features in different ranges without problems (bias to the larger values). Some features may have a value in the 1000-range and others in the 0-1 range.
$\begingroup$
$\endgroup$
2
-
3$\begingroup$ There are similar questions on StackOverflow (stackoverflow.com/q/8961586/1414455) and Quora (qr.ae/80zE4). The short answer is that you don't. $\endgroup$– tchakravartyNov 3, 2012 at 17:21
-
$\begingroup$ See stats.stackexchange.com/questions/255765/… $\endgroup$– FirebugAug 14, 2018 at 22:17
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
Partially answered in comments:
There are similar questions on StackOverflow (https://stackoverflow.com/questions/8961586/do-i-need-to-normalize-or-scale-data-for-randomforest-r-package) and Quora (https://www.quora.com/Machine-Learning/Should-inputs-to-random-forests-be-normalized?srid=3EJy&st=ns). The short answer is that you don't. – tchakravarty
answered Aug 14, 2018 at 21:57