Forest-based Regression - Output for Partial Dependence Plots?

[jang]

Hi,

I`m trying to run a forest-based regression in ArcGis Pro 3.0.3 with the Forest-based classification and regression tool (Spatial Statistics). I am searching for an output-option to create plots for marginal effects. In R, you would use Partial Dependence Plots (PDP) or Accumulated Local Effects (ALE) plots. Is there an option to create partial dependence plots with ArcGis Pro Tools?

The tool itself is great, as it lets you easily combine feature, raster and distance-based data, which you would have to manually create and combine in R. So it would be great, if I could use ArcGis for the whole process including plots.

Thank you for any insights or help on this.

[ShaunWalbridge · Accepted solution]

While we do have partial dependence plots for the presence-only prediction tools, we don't currently support those outputs for the Forest-based tools. The spatial statistics team has logged an issue on this, it's a good enhancement request! That said, if you'd like to do something like this in the short term, you may be served by using the ArcGIS bridge to connect between your analysis in ArcGIS Pro and those more fine grained plotting tools available in R. You can export your data directly to native R types like sf and raster objects, and then send it on to be plotted as you would with anything else within R. You can also create Geoprocessing scripts that call R code and integrate them into a broader workflow, albeit with some additional work required.

Cheers, Shaun