CAST: R Package for Spatial-Temporal Machine Learning Model Training

CAST is an R package providing supporting functionality to run the 'caret' package with spatial or spatial-temporal data. It includes methods for spatial cross-validation, variable selection, and estimating the area of applicability for prediction models, as described in multiple peer-reviewed publications from 2018 to 2024. The package was authored by Hanna Meyer.

Use Cases

• Estimating model performance for spatial predictions using the 'Nearest neighbor distance matching' cross-validation method.
• Selecting suitable predictor variables for spatial models based on their contribution to spatial model performance.
• Estimating the spatial area of applicability for prediction models to assess where predictions are reliable.

Strengths

• Methods are documented in multiple peer-reviewed publications spanning from 2018 to 2024.
• Specifically designed to address spatial and spatiotemporal modeling challenges within the established 'caret' framework.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and dataset scale are unknown, which may limit suitability assessment.
• License is listed as 'closed', which may restrict usage.

Provenance

Source

paperswithcode

Collection Method

Software package for R.

Time Range

Methods described in publications from 2018 to 2024.

Freshness

Last update date is unknown; freshness unverified.

Geography

Spatial coverage is not specified; package is designed for general spatial data.