- bug fix: default algorithm for FNN functions changed
- bug fix: tests run conditionally
- bug fix: fix failed tests in global_validation
- new features:
- calculate local point density within AOA
- option of spatial error profiles (errorProfiles with variable="geodist")
- normalize_DI for a more intuitive interpretation
- geodist allows calculating temporal distances
- ffs now can be run in parallel (Linux only)
- vignette "Cross-validation methods in CAST"
- knndm in feature space (experimental)
- nndm in feature space (experimental)
- modifications:
- function DItoErrormetric renamed to errorProfiles and allows for other dissimilarity measures
- Improvement and homogenization of plotting methods for nndm, knndm and geodist objects
- aoa and trainDI
weight
now allows list input - vignette on Introduction to CAST updated
- deprecated: *plot_geodist (replaced by plot.geodist) *plot_ffs (replaced by plot.ffs) *calibrate_aoa (replaced by errorProfiles)
- new features:
- CAST functions now return classes with generic plotting and printing
- new dataset for examples, tutorials and testing: data(splotdata)
- modifications:
- calibrate_aoa is now DItoErrormetric and returns a model (see function documentation)
- plot_geodist is now geodist. The result can be visualized with plot()
- plot_ffs is now plot(ffs)
- bug fix:
- fix issue #65 (threshold)
- deprecated (soon):
- plot_geodist, plot_ffs, calibrate_aoa
- bugfix:
- failed checks on Fedora 34 fixed
- new features:
- knndm as an alternative to nndm for large training data
- modifications:
- transition from raster to terra
- new features:
- Mahalanobis distance for AOA assessment as option
- modifications:
- faster estimation of the AOA
- parallel option for AOA deprecated (see vignette)
- delineation of the default threshold fixed as suggested in github.com//issues/46
- bugfix:
- fixed issue github.com/ropensci/rnaturalearth/issues/69
- new feature:
- nndm cross-validation as suggested by Milà et al. (2022)
- modifications
- plot_geodist works with NNDM
- trainDI works with NNDM
- rename of parameter folds in AOA and trainDI
- new feature:
- trainDI allows to calculate the DI of the training dataset separately from the aoa function
- plot and print functions for the AOA
- function to plot nearest neighbor distance distributions in geographic and feature space
- function global_validation added
- modifications
- extensive restructuring of the AOA function
- ffs and bss can be used with global_validation
- bugfix:
- error in manual assignment of weights fixed
- resolved dependence on package "GSIF" which was removed from the CRAN repository
- new feature:
- AOA can run in parallel
- calibration of the DI (calibrate_aoa)
- bugfix:
- aoa will work now with large training sets
- modifications:
- default threshold of AOA changed
- new feature:
- aoa now working with categorical variables
- bugfix:
- fixed error in ffs when >170 variables are used
- minor changes:
- changed order of parameters in aoa
- tutorial "Introduction to CAST" improved
- new feature:
- vignette: tutorial introducing the "area of applicability"
- variable threshold for aoa
- various modifications in aoa in line with submitted paper
- new feature:
- new function "aoa": quantify and visualize the area of applicability of spatial prediction models
- "minVar" in ffs: Instead of always starting with 2-pair combinations, ffs can now also be started with combinations of more variables (e.g starting with all combinations of 3)
- bugfix:
- ffs failed for "svmLinear" in previous version because of S4 class issues. Fixed now.
- bugfix:
- CreateSpaceTimeFolds accepts tibbles
- CreateSpaceTimeFolds automatically reduces k if necessary
- ffs accepts further arguments taken by caret::train
- new feature: plot_ffs has option to plot selected variables only
-
new feature: Best subset selection (bss) with target-oriented validation as (very slow but very reliable) alternative to ffs
-
minor adaptations: verbose option included, improved examples for ffs
-
bugfix: minor adaptations done for usage with plsr
-
new feature: Introduction to CAST is included as a vignette.
-
bugfix: minor error fixed in using user defined metrics for model selection.
-
bugfix:
ffs
with option withinSE=TRUE did not choose a model as "best model" if it was within the SE of a model that was trained in an earlier run but had the same number of variables. This bug is fixed and if withinSE=TRUE ffs now only compares the performance to models that use less variables (e.g. if a model using 5 variables is better than a model using 4 variables but still in the SE of the 4-variable model, then the 4-variable model is rated as the better model). -
new feature:
plot_ffs
plots the results of ffs to visualize how the performance changes according to model run and the number of variables being used.
Initial public version on CRAN