How to balance classes in H2O's Auto ML function?

user_1823293 0
Viewed 858

could someone please let me know if it is possible to specify methods to balance classes while using H2O's Auto ML function for classificatin problems? 

h2o.automl(x, y, training_frame, validation_frame = NULL,
leaderboard_frame = NULL, nfolds = 5, fold_column = NULL,
weights_column = NULL, max_runtime_secs = 3600, max_models = NULL,
stopping_metric = c("AUTO", "deviance", "logloss", "MSE", "RMSE", "MAE",
"RMSLE", "AUC", "lift_top_group", "misclassification",
"mean_per_class_error"), stopping_tolerance = NULL, stopping_rounds = 3,
seed = NULL, project_name = NULL)

The existing documentation doesn't seem to specify any parameters. Thanks.

h2o machine-learning r
1 Answers

The balance_classes option is not exposed to the user (as of H2O v3.16.0.1) in the AutoML function, however we have a ticket open to turn on automatic class-balancing when the response is reasonably imbalanced (e.g. <10% minority class). This should be added soon.

user_54513440
Related