search

semiautomaticgit / SemiAutomaticClassificationPlugin

https://fromgistors.blogspot.com/p/semi-automatic-classification-plugin.html
Other
136 stars 50 forks source link

Regressions in SCP? #306

Open jgrn307 opened 8 months ago

jgrn307 commented 8 months ago

Is there currently support for continuous variable-type analyses in SCP? I know a lot of the various ML techniques like random forest support continuous variables. It seems the workflow shouldn't be too hard to adapt, if there isn't -- use the ROI tool but have a column for continuous variables instead of class variables when doing the extraction. Thoughts?

semiautomaticgit commented 8 months ago

@jgrn307 thank you, I'll consider adding ML regression to the algorithms. Probably the input should be already a raster with continuous variables, because I think the ROI tool should be specific for thematic classifications.

jgrn307 commented 8 months ago

So the training data (polygons or points) would contain the response variable as attributes, and the spectra would be extracted and "joined" to those points for a model to be built. ROIs are probably not the right term, exactly -- just vector data + attributes containing a continuous variable.

semiautomaticgit commented 8 months ago

So the training data (polygons or points) would contain the response variable as attributes, and the spectra would be extracted and "joined" to those points for a model to be built. ROIs are probably not the right term, exactly -- just vector data + attributes containing a continuous variable.

Yes something like this, or a rasterized vector