LEGION-package: LEGION - a tool for computing several artificial layers (AL)

In SchoenbergA/LEGION: artifical layer computing tool

Description

In general the LEGION package provides tools to compute a huge stack of artifical layers to use in maschine learing or ordination/cluster analysis. These methods require lots of information, the more the better (unless it leads to overfitting). With LEGION the user can first compute several indices and apply filter to generate artifical layers of a RasterLayer or a RasterStack. Further LEGION provides a function to detect higly correlating layers (which could lead to overfitting and/or high computing time while processing).

Details

LEGION

Note

In general a 'brute force' use of LEGION is not recommended. E.g. applying all available filters with multiple sizes of MovingWindows on all availabe indices would lead to multiple hundreds of artifal layers. This could take hours to days to compute (especially in the case of large RasterStacks) which makes little sense.

On the other hand LEGION helps to reduce "overthinking" in planning which ALs should be used for a maschine learing or ordination/cluster analysis task. A recommended way to generate a RasterStack with the required information but minimal correlation in a relatively short time is to first generate the indices and then to clean the RasterStack from correlating ALs and then successively to filter the cleaned RasterStack.

Author(s)

Andreas Schönberg

SchoenbergA/LEGION documentation built on Jan. 31, 2021, 10:12 a.m.