Xtestmerger: merge training set (X) and (test) set

In sorhawell/forestFloor: Visualizes Random Forests with Feature Contributions

Description

... and expand inbag matrix and training target vector to compute FC for a test set.

Usage

Xtestmerger(X,test,inbag=NULL,y=NULL)

Arguments

X	X , training set data.frame used to train a random forest model
test	test, a test set data.frame which feature contributions should be computed for
inbag	matrix of inbag sampling to expande with training set, which is set OOB for any tree
y	random forest target vector, which is set to first value for observation

Details

Xtestmerger is a low-level function to merge a test set with X training set. There can be no names, column class, column number mismatch. Moreover any level in any factor of test must be present in X, as RF/forestFloor cannot score a unknown factor level / category.

Value

List of merged bigX, bigInbag and bigy. The two latter may be NULL if not provided.

Author(s)

Soren Havelund Welling

Examples

library(randomForest)
library(forestFloor)
#X y could be a training set
X = data.frame(numeric = c(1,5,2,7,-4.3),
               factor1 = factor(c("jim","freddy","marley","marley","alfred")),
               factor2 = factor(c("jill","ann","liz","leila","vicky")))
y = factor(1:5)
set.seed(1)
rf = randomForest(X,y,keep.inbag=TRUE,ntree=7)
#should not raise any error
test = data.frame(numeric = rnorm(5),
                  factor1 = factor(c("jim","jim","jim","freddy","freddy")),
                  factor2 = factor(c("jill","jill","vicky","leila","vicky"))
                  )
out = Xtestmerger(X,test,inbag=rf$inbag,y=y)

sorhawell/forestFloor documentation built on Oct. 23, 2021, 2:20 a.m.