Hi!
I have a question regarding Random Trees in ArcMap.
I am running “train random trees” and then using the output as input in the “classify raster” tool.
When I rerun the process above, using the same parameters in the tools, I get a different output. (I am setting the “max_samples_per_class” to 0 in order to make sure all my training data is used.)
I am wondering why this is and have a couple of possible ideas;
1. Is a portion of the training data randomly set aside each time to calculate the variable importance, oob error and etc? If so, how much of the data is set aside?
2. Is the output different each time because the choice of which predictorvariable (choice of which 2 bands in the input image that should be compared) to consider at each tree node is randomized?
I am struggling with understanding what the tool actually does behind the scenes?
Any inputs or ideas are much appreciated
Thank you!
