I see h2o model performance metric contains AUC, logloss etc. There is one model performance metric called lift_top_group, is it lift on top decile?
Also can user specify the band for h2o to output gains chart such as top 5%, 5%-10%, 10%-15% ....... The function I can find is h2o.gainsLift
You can see all quantile groups in the output (Flow has a nice display). The top group is the top 1%, and lift_top_group refers to that. It can be used for early stopping. All other information from the gains/lift chart is available in all binomial model metrics for inspection. More info here: https://github.com/h2oai/h2o-3/blob/master/h2o-docs/src/product/tutorials/GainsLift.md (note that the picture is outdated, it uses 5% intervals uniformly - now there are more fine grain groups at the top).