rfokrfidwcv: Cross validation, n-fold for the average of the hybrid method...

In spm: Spatial Predictive Modeling

Cross validation, n-fold for the average of the hybrid method of random forest and ordinary kriging and the hybrid method of random forest and inverse distance weighting (RFOKRFIDW)

Description

This function is a cross validation function for the average of the hybrid method of random forest and ordinary kriging and the hybrid method of random forest and inverse distance weighting (RFOKRFIDW).

Usage

rfokrfidwcv(
  longlat,
  trainx,
  trainy,
  cv.fold = 10,
  mtry = function(p) max(1, floor(sqrt(p))),
  ntree = 500,
  idp = 2,
  nmaxok = 12,
  nmaxidw = 12,
  vgm.args = ("Sph"),
  block = 0,
  predacc = "VEcv",
  ...
)

Arguments

longlat	a dataframe contains longitude and latitude of point samples (i.e., trainx and trainy).
trainx	a dataframe or matrix contains columns of predictive variables.
trainy	a vector of response, must have length equal to the number of rows in trainx.
cv.fold	integer; number of folds in the cross-validation. if > 1, then apply n-fold cross validation; the default is 10, i.e., 10-fold cross validation that is recommended.
mtry	a function of number of remaining predictor variables to use as the mtry parameter in the randomForest call.
ntree	number of trees to grow. This should not be set to too small a number, to ensure that every input row gets predicted at least a few times. By default, 500 is used.
idp	numeric; specify the inverse distance weighting power.
nmaxok	for local predicting: the number of nearest observations that should be used for a prediction or simulation, where nearest is defined in terms of the space of the spatial locations. By default, 12 observations are used for OK.
nmaxidw	for local predicting: the number of nearest observations that should be used for a prediction or simulation, where nearest is defined in terms of the space of the spatial locations. By default, 12 observations are used for IDW.
vgm.args	arguments for vgm, e.g. variogram model of response variable and anisotropy parameters. see notes vgm in gstat for details. By default, "Sph" is used.
block	block size. see krige in gstat for details.
predacc	can be either "VEcv" for vecv or "ALL" for all measures in function pred.acc.
...	other arguments passed on to randomForest or gstat.

Value

A list with the following components: for numerical data: me, rme, mae, rmae, mse, rmse, rrmse, vecv and e1; or vecv.

Note

This function is largely based on rf.cv (see Li et al. 2013) and rfcv in randomForest. When 'A zero or negative range was fitted to variogram' occurs, to allow gstat running, the range was set to be positive by using min(vgm1$dist). In this case, caution should be taken in applying this method, although sometimes it can still outperform IDW and OK.

Author(s)

Jin Li

References

Li, J. 2013. Predicting the spatial distribution of seabed gravel content using random forest, spatial interpolation methods and their hybrid methods. Pages 394-400 The International Congress on Modelling and Simulation (MODSIM) 2013, Adelaide.

Liaw, A. and M. Wiener (2002). Classification and Regression by randomForest. R News 2(3), 18-22.

Examples

## Not run: 
data(petrel)

rfokrfidwcv1 <- rfokrfidwcv(petrel[, c(1,2)], petrel[, c(1,2, 6:9)], petrel[, 5],
predacc = "ALL")
rfokrfidwcv1

n <- 20 # number of iterations, 60 to 100 is recommended.
VEcv <- NULL
for (i in 1:n) {
rfokrfidwcv1 <- rfokrfidwcv(petrel[, c(1,2)], petrel[, c(1,2,6:9)], petrel[, 5],
predacc = "VEcv")
VEcv [i] <- rfokrfidwcv1
}
plot(VEcv ~ c(1:n), xlab = "Iteration for RFOKRFIDW", ylab = "VEcv (%)")
points(cumsum(VEcv) / c(1:n) ~ c(1:n), col = 2)
abline(h = mean(VEcv), col = 'blue', lwd = 2)

n <- 20 # number of iterations, 60 to 100 is recommended.
measures <- NULL
for (i in 1:n) {
rfokrfidwcv1 <- rfokrfidwcv(petrel[, c(1,2)], petrel[, c(1,2,6:9)], petrel[, 5],
predacc = "ALL")
measures <- rbind(measures, rfokrfidwcv1$vecv)
}
plot(measures ~ c(1:n), xlab = "Iteration for RFOKRFIDW", ylab = "VEcv (%)")
points(cumsum(measures) / c(1:n) ~ c(1:n), col = 2)
abline(h = mean(measures), col = 'blue', lwd = 2)

## End(Not run)

spm documentation built on May 6, 2022, 9:06 a.m.