The m=√p rule for random forests

A couple of days ago, in our lab session, we discussed random forrests, and, since it was based on the example in ISLR, we had a quick discussion about the random choice of features, and the “” rule Interestingly, on that one, we can play a bit, and try all choices, and do it again, on a different train/test split, library(randomForest) library(ISLR2) set.seed(123) sim = function(t){ train = sample(nrow(Boston), size = nrow(Boston)*.7) subsim = function(i){ rf.boston <- randomForest(medv ~ ., data = Boston, subset … <a href=“https://freakonometrics.hypotheses.org/77919" …