Trainable Unit Properties

Field Description

Field Name

Name. Name of the trainable unit. It should satisfy context naming conventions. The name is required to refer to this trainable unit from other parts of the system.

name

Description. Textual description of the trainable unit.

description

Task. Machine learning task, Regression, Classification or Anomaly Detection. Once a trainable unit is created, the task cannot be changed for this trainable unit.

task

Algorithm. Machine learning algorithm to solve a specific business problem. The list of available algorithms depends on the chosen task. See Algorithms for more detail.

algorithm

Hyperparameters. Parameters of the chosen algorithm that are not learned during the training process. See Algorithm Hyperparameters for more detail.

hyperparameters

Label Column Field Name. Field name of the label column (target variable). See Arguments of trainable unit functions for more detail.

labelFieldName

Dataset Has Weights. Determines whether or not dataset instances are weighted.

hasWeights

Weight Field Name. Name of the field containing weights if dataset instances are weighted. Visible only if the Dataset Has Weights option is set to true.

weightFieldName

Number of Folds. The number of folds into which the dataset will be partitioned for cross-validation (when the crossValidate function is called).

cvNumFolds

Random Number Seed. The seed number for the random number generator used for partitioning the data into folds for cross-validation.

cvRandomSeed