score_tree_interval

  • Available in: GBM, DRF, XGBoost
  • Hyperparameter: no

Description

The score_tree_interval option specifies to score the model after this many trees. This option is useful when used with early stopping and attempting to make early stopping reproducible. It can be disabled by setting this value to 0.

When specifying early stopping parameters, the stopping_rounds option applies to the number of scoring intervals or iterations (with score_each_iteration) H2O has performed, so regular scoring intervals of small size help control early stopping the most (though there is a speed tradeoff to scoring more often). The default is to use H2O’s assessment of a reasonable ratio of training time to scoring time, which often results in inconsistent scoring gaps.

Example

library(h2o)
h2o.init()

# import the cars dataset:
# this dataset is used to classify whether or not a car is economical based on
# the car's displacement, power, weight, and acceleration, and the year it was made
cars <- h2o.importFile("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")

# convert response column to a factor
cars["economy_20mpg"] <- as.factor(cars["economy_20mpg"])

# set the predictor names and the response column name
predictors <- c("displacement","power","weight","acceleration","year")
response <- "economy_20mpg"

# split into train and validation sets
cars.split <- h2o.splitFrame(data = cars,ratios = 0.8, seed = 1234)
train <- cars.split[[1]]
valid <- cars.split[[2]]

# try using the `score_tree_interval` parameter:
cars_gbm <- h2o.gbm(x = predictors, y = response, training_frame = train,
                    validation_frame = valid, score_tree_interval = 5, seed = 1234)

# print the model score every 5 trees
print(h2o.scoreHistory(cars_gbm))

import h2o
from h2o.estimators.gbm import H2OGradientBoostingEstimator
h2o.init()

# import the cars dataset:
# this dataset is used to classify whether or not a car is economical based on
# the car's displacement, power, weight, and acceleration, and the year it was made
cars = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/junit/cars_20mpg.csv")

# convert response column to a factor
cars["economy_20mpg"] = cars["economy_20mpg"].asfactor()

# set the predictor names and the response column name
predictors = ["displacement","power","weight","acceleration","year"]
response = "economy_20mpg"

# split into train and validation sets
train, valid = cars.split_frame(ratios = [.8], seed = 1234)

# try turning on the `score_tree_interval` parameter:
# initialize your estimator
cars_gbm = H2OGradientBoostingEstimator(score_tree_interval = 5, seed = 1234)

# then train your model
cars_gbm.train(x = predictors, y = response, training_frame = train, validation_frame = valid)

# print the model score every 5 trees
cars_gbm.scoring_history()