I have been looking at the documentation for the Train Random Trees Classifier for Desktop 10.7 and the documentation for the Forest-based Classification and Regression tools in Pro and have a couple of questions that I hope someone may be able to answer. There is more documentation on the RF tool for Pro, however I am trying to do this on Desktop as that is what I have access to, so am just wondering how much of the information on the Pro tool applies to the Desktop version?
For example, my main query relates to this. The RF tool on Pro states that
By default, 10 percent of the training data is excluded from training for validation purposes. After the model is trained, it is used to predict the values for the test data, and those predicted values are compared to the observed values to provide a measure of prediction accuracy based on data that was not included in the training process. Additional diagnostics about the model, including forest characteristics, Out of Bag (OOB) Errors, and a Summary of Variable Importance are also included.
Does this also apply for the RF classifier on Desktop 10.7? Is 10% of the training data also set aside and is this how the training accuracy value that shows up in the Geoprocessing results is calculated?
And is there any way to get the OOB errors out of the Desktop tool? Or is that only available using Pro?
Appreciate any help!
Don't know... but if the tool has a history and the Spatial Statistics extension is the same, perhaps, examine one run, because maybe not all the information is revealed during a run... worth a shot anyway
You can also access the messages for a previous run of the Forest-based Classification and Predictiontool via the Geoprocessing history. The messages include information on the characteristics of your model, OOB errors, variable importance, and validation diagnostics.
and some more information for slightly different versions of both