How does h2o.predict calculate the probabilities of different classes ? In the randomForest package of R, the probabilities are calculated based on the number of votes. For example probability of being class B = number of votes for class B / total number of trees. I can't seem to understand how h2o's RF algorithm provides so much granular probabilities even with few trees
Best Answer
According to this Google group Random Forest of H20 calculates probabilities at each leaf rather than calculating a vote. And the prediction is the average of probabilities instead of average of votes.
And according to this benchmark the Random Forest of H20 is more accurate than classical Random Forest (randomForest function in R and Random Forest in scikit-learn in Python).
Similar Posts:
- Solved – Random Forest probability
- Solved – Random forest “certainty / likelihood score” – how to score records in RF mode in R?
- Solved – How to make the randomforest trees vote decimals but not binary
- Solved – Scikit Random forest pred_proba gives rounded off values
- Solved – Predict function tuning for random forest