I am a basic R user. Could you please help me on how to apply Spatial Cross-Validation using mlr3 or another library such as random forest. I want to use 6 distinct raster data for this study.
I am trying to use random forest algorithm, tune the model parameters, evaluate and predict the result on all data.
- Landsat 4 tiles as a reference data to train and test my model (expected out put).
- MODIS surface reflectance 1-7 bands, NDVI, NDWI, DEM as input variables.
- MODIS land cover for comparison with the model out put.
I tried the following code lines to define my input data but I did create a regression task.
library(raster)
library(mlr3)
library(mlr3spatiotempcv)
library(mlr3tuning)
v1 <- list.files("C:/Users/CTE/Desktop/variable_1") # Landsat data to train and test RF model
v2 <- list.files("C:/Users/CTE/Desktop/variable_2") # MODIS 1-7 bands
v3 <- list.files("C:/Users/CTE/Desktop/variable_3") # NDVI
v4 <- list.files("C:/Users/CTE/Desktop/variable_4") # NDWI
v5 <- ("C:/Users/CTE/Desktop/variable_5") # DEM
v6 <- list.files("C:/Users/CTE/Desktop/variable_6") # MODIS LC for comparison with the result of RF model
task = mlr3spatiotempcv::as_task_regr_st(dplyr::select( ),
id = "", target = "") #Create a regression task.
I look at some example codes but I did not find any similar code to solve this problem.
Any help is highly appreciated. Looking forward to your kind assistance.