Closed mralbu closed 3 years ago
This sounds interesting! I'll have a look at this, when I've got some spare time. @mralbu : Can you open a PR with this branch? Even if this is still work in progress, this would be the best place to discuss details ;-)
Thanks for your support! Seb
Opened pull request #165.
Merged into develop.
Thanks for this great package! I was introduced to regression kriging reading the documentation and found it super useful. I've created a ClassificationKriging class adapting RegressionKriging to work with categorical variables, by mapping the scikit-learn classifiers predict_proba simplexes to an isometric space using the ilr transform, and kriging/interpolating the residuals of the (also ilr transformed) observed data. I am far from being an expert on geostatistics or compositional data analysis, but I believe this method corresponds to the Simplicial Indicator Kriging method. Do you think this method could be a good addition to PyKrige? Initial implementation avaiable in the classification-kriging branch.