Spatial and Environmental Blocking for K-Fold and LOO Cross-Validation
blockCV: Spatial and Environmental Blocking for K-Fold and LOO Cross-V...
Use distance (buffer) around records to separate train and test folds
Explore spatial block size
Use buffer around records to separate train and test folds (a.k.a. buf...
Use environmental or spatial clustering to separate train and test fol...
Use the Nearest Neighbour Distance Matching (NNDM) to separate train a...
Visualising folds created by blockCV in ggplot
Compute similarity measures to evaluate possible extrapolation in test...
Measure spatial autocorrelation in spatial response data or predictor ...
Use spatial blocks to separate train and test folds
Use environmental clustering to separate train and test folds
Explore the generated folds
Explore spatial block size
Measure spatial autocorrelation in the predictor raster files
Use spatial blocks to separate train and test folds
Creating spatially or environmentally separated folds for cross-validation to provide a robust error estimation in spatially structured environments; Investigating and visualising the effective range of spatial autocorrelation in continuous raster covariates and point samples to find an initial realistic distance band to separate training and testing datasets spatially described in Valavi, R. et al. (2019) <doi:10.1111/2041-210X.13107>.