[R] RandomForest and Missing Values
Lorenzo Isella
lorenzo.isella at gmail.com
Mon Jan 28 17:37:25 CET 2013
Dear All,
I would like to use a randomForest algorithm on a dataset.
The set is not particularly large/difficult to handle, but it has some
missing values (both factors and numerical values).
According to what I found
https://stat.ethz.ch/pipermail/r-help/2005-September/078880.html
https://stat.ethz.ch/pipermail/r-help/2007-January/123117.html
the randomForest package has a problem with missing data (essentially
you have to resort to some "trick" to introduce them into your dataset
--a median value, the most common factor, a linear interpolation
etc...).
Seen that I could not find a clear workaround for this (but I cannot
be the only one who has in mind to do a randomForest on a less than
perfect data set), can anyone help me out?
I am concerned about the consequences of introducing the missing
values into the data set.
The cforest function in the "Party" package does not seem to have this
limitation, but on the other hand the randomForest package has passed
the test of time....so should I drop it in this case?
Any suggestion is appreciated.
Cheers
Lorenzo
More information about the R-help
mailing list