It's worth mentioning that with a large number of data points, each tree is trained on 1 - 1/e (~63%) of the training data on average. So each data point will tend to be used in training for 3 out of every 5 trees, and be OOB for 2 in 5 trees. So the OOB score does utilize a large chunk of the overall ensemble.
The proof is left as an exercise for the reader :)
It's worth mentioning that with a large number of data points, each tree is trained on 1 - 1/e (~63%) of the training data on average. So each data point will tend to be used in training for 3 out of every 5 trees, and be OOB for 2 in 5 trees. So the OOB score does utilize a large chunk of the overall ensemble.
The proof is left as an exercise for the reader :)
Thanks for adding that, Joe. Been some time since I heard from you, happy to see you adding value here :)
you are good, is there something climate analysis and forecast related?