2
$\begingroup$

When a random forest model is built, does the out-of-bag error consider all the data points in the set, at the end?

I know that it is basically asking decision trees to make predictions on data points that was not part of their training data. However when this exercise is completed, will all the data points in the set be accounted for?

$\endgroup$
0

1 Answer 1

2
$\begingroup$

A forest consists of a large number of trees. A random forests consists of a large number of decision or regression trees. Each of these trees will be grown with a different random set of the data. Thus with a sufficiently large number of trees all the data will be accounted for in the construction of the forest but not in the construction of each tree. Obviously there is a chance of a single datapoint not being represented in any of the trees, but every tree will on average contain about two thirds of all the data (cd https://stats.stackexchange.com/a/88994/117812 ) and the preset for ntree in the randomforest package in the randomForest R package is 500.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.