CODEWITHSUNDEEP
pythonscikit-learnregressionrandom-forest
Enrique G. Ortiz
Enrique G. Ortiz

Reputation: 1

Leaf Indices Off for scinkit-learn Random Forest Regression

I am trying to use scinkit-learn's apply function for the RandomForestTreeRegressor to obtain the leaf indices for each learned tree for some data. I have specified a max_depth of 3, which should result in a maximum leaf index of 4, but I am actually getting number much higher indices in the 40s. Is there any explanation for this?

I figured maybe something was wrong with my data, so I tried the same with the sample code on scikit's page and found the same issue. If you want to reproduce it, you can edit the source code here: http://bit.ly/1GHz1iG

Upvotes: 0

Views: 316

Answers (1)

Andreas Mueller
Andreas Mueller

Reputation: 28768

The indices actually go over all nodes in the tree, not only the leaves. There are up to 2 ** max_depth leaves (which would be 8 in your case, not sure why you expect 4). As all nodes are numbered, though, the numbers can go up to 2 ** (max_depth + 1) - 1

Upvotes: 1

Related Questions