Getting prediction intervals from GBM models

Question

I am working with GBM regression models(in H2O) and am using Quantile distribution for the distribution parameter. I am looking for a method to provide prediction intervals in addition to point value prediction. What is the best way to achieve this generally?

Since we are using quantile regression, I was hoping we should be able to exploit it as follows: Lets say we are interested to find 95% prediction interval. We train one GBM model with quantile alpha=.025 and another with .975. And then we call predict on these 2 models to get lower and upper values of the range. Does this approach seem correct?

Answer 1

Yes. H2O's implementation seems like a more robust, distributed version of the GBM model offered by sklearn. If you look into the documentation provided by sklearn they offer the following option for the loss function:

Quantile ('quantile'): A loss function for quantile regression. Use 0 < alpha < 1 to specify the quantile. This loss function can be used to create prediction intervals

They also offer an example on how quantile regression can be used to create prediction intervals.