Supporting functionality to run 'caret' with spatial or spatial-temporal data. 'caret' is a frequently used package for model training and prediction using machine learning. This package includes functions to improve spatial-temporal modelling tasks using 'caret'. It prepares data for Leave-Location-Out and Leave-Time-Out cross-validation which are target-oriented validation strategies for spatial-temporal models. To decrease overfitting and improve model performances, the package implements a forward feature selection that selects suitable predictor variables in view to their contribution to the target-oriented performance.
| Version: |
0.3.1 |
| Depends: |
R (≥ 3.1.0) |
| Imports: |
caret, stats, utils, ggplot2, graphics |
| Suggests: |
doParallel, GSIF, randomForest, lubridate, raster, sp, knitr, mapview, rmarkdown |
| Published: |
2018-11-19 |
| Author: |
Hanna Meyer [cre, aut],
Chris Reudenbach [ctb],
Marvin Ludwig [ctb],
Thomas Nauss [ctb] |
| Maintainer: |
Hanna Meyer <hanna.meyer at geo.uni-marburg.de> |
| License: |
GPL (≥ 3) | file LICENSE |
| URL: |
https://github.com/environmentalinformatics-marburg/CAST |
| NeedsCompilation: |
no |
| Materials: |
README NEWS |
| CRAN checks: |
CAST results |