← All types

MachineLearningServerDefaults

The machine-learning values this server uses for an instance that has not set its own.

Fields

Whether system models are trained.

Hours between retraining runs.

Finished operations needed before a first model is trained. System models also wait for enough to hold out 30 for evaluation (150 at the default 20% hold-out).

New finished operations needed before a model is retrained.

minR2

Float!Non-null

Lowest R² a regression model may have to be used.

maxRMSE

FloatNullable

Highest RMSE a regression model may have to be used; null for no limit.

Lowest accuracy a classification model may have to be used.

Referenced by (1)
Exact type SDL
"""
The machine-learning values this server uses for an instance that has not set its own.
"""
type MachineLearningServerDefaults {
  """Whether system models are trained."""
  enabled: Boolean!
  """Hours between retraining runs."""
  retrainIntervalHours: Int!
  """
  Finished operations needed before a first model is trained. System models also wait for enough to hold out 30 for evaluation (150 at the default 20% hold-out).
  """
  minTrainingRows: Int!
  """New finished operations needed before a model is retrained."""
  retrainThresholdRows: Int!
  """Lowest R² a regression model may have to be used."""
  minR2: Float!
  """
  Highest RMSE a regression model may have to be used; null for no limit.
  """
  maxRMSE: Float
  """Lowest accuracy a classification model may have to be used."""
  minAccuracy: Float!
}