buildRanger: ranger Interface

In bartzbeielstein/SPOT: Sequential Parameter Optimization Toolbox

Description

This is a simple wrapper for the ranger function from the ranger package. The purpose of this function is to provide an interface as required by SPOT, to enable modeling and model-based optimization with ranger.

Usage

buildRanger(x, y, control = list())

Arguments

x	matrix of input parameters. Rows for each point, columns for each parameter.
y	one column matrix of observations to be modeled.
control	list of control parameters. These are all configuration parameters of the ranger function, and will be passed on to it.

Value

an object of class "spotRanger", with a predict method and a print method.

Examples

## Not run: 
## Create a simple training data set
testfun <- function (x) x[1]^2
x <- cbind(sort(runif(30)*2-1))
y <- as.matrix(apply(x,1,testfun))
## test data:
xt <- cbind(sort(runif(3000)*2-1))
## Example with default model (standard randomforest)
fit <- buildRanger(x,y)
yt <- predict(fit,data.frame(x=xt))
plot(xt,yt$y,type="l")
points(x,y,col="red",pch=20)
## Example with extratrees, an interpolating model
fit <- buildRanger(x,y,
                   control=list(rangerArguments = 
                                list(replace = F,
                                   sample.fraction=1,
                                   min.node.size = 1,
                                   splitrule = "extratrees")))
yt <- predict(fit,data.frame(x=xt))
plot(xt,yt$y,type="l")
points(x,y,col="red",pch=20)

## End(Not run)

bartzbeielstein/SPOT documentation built on June 13, 2020, 5:58 p.m.