[R-pkgs] randomForestSRC 2.9.0 is now available
Udaya B. Kogalur
kog@|ur@he@r @end|ng |rom gm@||@com
Mon Apr 22 20:53:14 CEST 2019
It's been some time since we last sent out an announcement, so this one
will cover more than just the last update.
The latest release of randomForestSRC is now available on CRAN at:
The GitHub repository, through which we prefer to receive bug reports, is
If you do find issues, please use:
and take the time to post a minimal script (and data set if necessary) that
isolates the error.
Additional documentation can be found at:
Details are as follows:
Ensembles in regression now support Greenwald-Khanna approximate quantile
queries via rfsrc(), predict.rfsrc() and the new wrapper
quantileReg.rfsrc(). Related to this, a new split rule "quantile.regr" has
Another new wrapper, imbalanced.rfsrc(), implements various solutions to
the two-class imbalanced problem, including the newly proposed
quantile-classifier approach of O'Brien and Ishwaran (2017). This also
includes Breiman's balanced random forests under-sampling of the majority
class. Performance is assessed using the G-mean, but misclassification
error can be requested.
Also, the new parameter get.tree in predict.rfsrc() allows users to extract
the ensembles for a single tree or subset of trees over the forest.
The default nodesize for survival and competing risk has been changed to 15.
We've added new splitrules "auc" and "entropy" for classification. A new
variable importance methodology called Holdout VIMP has been implemented.
Here, we exclude a variable from a subset of trees and compare the error
rates between those trees in which the variables was included against those
in which it was excluded. The key point here is that no permutation of a
variable is conducted. See holdout.vimp.rfsrc() and the associated Rd file
for more information.
Finally, some function names were changed as a general move towards name
uniformity in the package. Sorry about that.
Unfortunately there has been no further work on the Spark build. However,
the Java wrappers have been kept up to date, and the Hello World script is
still functional. Instructions are provided here:
For those requesting and still awaiting CPU performance enhancements, our
continued apologies. Our focus has been methodological, but our intention
is to address performance in the next build. We promise.
Udaya B. Kogalur, Ph.D.
ubkogalur using gmail.com
[[alternative HTML version deleted]]
More information about the R-packages