check_constant_response

  • Available in: GBM, DRF, Uplift DRF

  • Hyperparameter: no

Description

This option checks if a response column is a constant value. If this option is enabled (default), then an exception is thrown if the response column is a constant value. If this option is disabled, then the model will train regardless of the response column being a constant value or not.

Example

library(h2o)
h2o.init()

# import the iris dataset:
train <- h2o.importFile("https://s3.amazonaws.com/h2o-public-test-data/smalldata/iris/iris_train.csv")
train$constantCol <- 1

# Build a GBM model. This should run successfully when
# check_constant_response is set to false.
iris_gbm_initial <- h2o.gbm(y = 6, x = 1:5, training_frame = train)
import h2o
from h2o.estimators.gbm import H2OGradientBoostingEstimator
h2o.init()

# import the iris dataset:
train = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/iris/iris_train.csv")
train["constantCol"] = 1

# Build a GBM model. This should run successfully when
# check_constant_response is set to false.
my_gbm = H2OGradientBoostingEstimator(check_constant_response=False)
my_gbm.train(x=list(range(1,5)), y="constantCol", training_frame=train)