0
$\begingroup$

I have used a random forest model for prediction of prices. Should the model be predictable in its behavior? By this, I mean that I'm not changing the model and the input, Will the predicted value always be the same?

$\endgroup$

2 Answers 2

3
$\begingroup$

Yes, the output is the same for the same input.

The "randomness" just concerns the approach how an ensemble is built: bagging (considering random subset of objects) and random selection of features for splitting (when we create each decision tree and choose which feature would be used as a splitting criteria).

$\endgroup$
1
  • $\begingroup$ Exactly. Once the model has been fit, the model will be persistent, as the branches of each estimator and thresholds have been set and each sample will fall into the exact same leaf $\endgroup$
    – yatu
    Sep 6, 2019 at 11:56
1
$\begingroup$

Yes. Prediction result of Random Forest model is an average of results of all individual trees, there is no randomness, since you are not changing the trees and result per tree is deterministic.

As, Lana said, randomness is only involved when Random Forest model is being trained.

$\endgroup$

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.