LLTO_cv: Leave Location Time Out Cross Validation

In kanj241/agrosoil: Surface and Rootzone Soil Moisture Prediction

Leave Location Time Out Cross Validation

Description

Assess model performance (R2, RMSE, ME) via a Leave Location Time Out Cross validation algorithm (LLTO-CV).

Usage

LLTO_cv(
  log_ = FALSE,
  targ_var,
  reg_mat_,
  pnts_cv,
  idcol,
  nfold,
  fmula,
  kan,
  month.lst_,
  pred.year_,
  method = "ranger"
)

Arguments

log_	TRUE/FALSE. Whether to noramalise observed and predicted soil moisture measurements during RMSE estimation.
targ_var	Predicted target variable
reg_mat_	Regression matrix for cross validation
pnts_cv	Object of class SpatialPoints dataframe
idcol	Column name / Station ID of in situ measurement
nfold	Numeric. Number of cross validation folds
fmula	Regression formula
kan	Character. Single month of the targetted soil moisture prediction year. Values are abbreviated month name. E.g. "Jan", "Mar", "Oct".
month.lst_	Vector or a list. All months of the targetted soil moisture prediction year.
pred.year_	Numeric. Single soil moisture prediction year
method	Algorithm for LLTO. Default value is "ranger". Other option is "ensemble" for coupling Random forest (ranger) and Gradient boost (Xgboost) algorithms.

Value

A list with CV resutls (Observed, predicted and performance metrics)

Examples

library(raster)
library(caret)
library(ranger)
library(DescTools)

out <- LLTO_cv(log_, targ_var, reg_mat_, pnts_cv, idcol, nfold, fmula, kan, month.lst_, pred.year_, method)

kanj241/agrosoil documentation built on March 25, 2022, 12:22 a.m.