Suggests Model Parameters

Provides suggestions of the model parameters, namely time, interpretability, and accuracy.

dai.suggest_model_params(
  training_frame,
  target_col,
  is_classification,
  is_timeseries,
  is_image,
  config_overrides = "",
  cols_to_drop = NULL
)

Arguments

training_frame

DAIFrame to be used for the model training.
target_col

The name of the target variable.
is_classification

Whether the predicted variable is categorical (true) or numerical (false).
is_timeseries

Whether the target variable is a time-series or not.
is_image

Whether predictor variables contain images
config_overrides

Raw config.toml file content (UTF8-encoded string) or a list/vector of config variables.
cols_to_drop

A character vector of column names to be dropped from the data (optional).

Value

A list with the suggested parameter values.

See also

dai.train, dai.preview_experiment

Examples

dai.connect(uri = 'http://127.0.0.1:12345', username = 'h2oai', password = 'h2oai')
iris_dai <- as.DAIFrame(iris, progress = FALSE)
suggested_params <- dai.suggest_model_params(training_frame = iris_dai,
                                             target_col = 'Species',
                                             is_classification = TRUE,
                                             is_timeseries = FALSE,
                                             is_image = FALSE)
print(suggested_params)
suggested_params$progress <- FALSE
# train model with suggested parameters
model <- do.call(dai.train, suggested_params)