Landmap package - automated mapping using Ensemble Machine Learning

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

Landmap package - automated mapping using Ensemble Machine Learning

Tomislav Hengl-5

I've wrapped functions extending #SuperLearner and #subsemble R pkgs
(Ensemble Machine Learning) which now can be used to generate Spatial
Predictions (fully automated framework, requires no special
geostatistical assumptions or choices). You can install the working
package from:

https://github.com/Envirometrix/landmap

The general method is explained in https://peerj.com/articles/5518/

The package can also be used to download some of the 2TB GeoTiffs
available via https://openlandmap.org.

Contributions and comments are welcome.

Few things I am still struggling with are:

1. Deriving prediction errors and predictions intervals
(https://github.com/Envirometrix/landmap/blob/master/R/train.spLearner.R#L252),
2. Generating spatial simulations using EML models,
3. Building models with larger data sets e.g. >>1000 points,

Thanks in advance,

Tom Hengl
https://opengeohub.org/people/tom-hengl

_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo