Gradient-Boosted Tree Regression

This operator implements the Gradient-Boosted Tree Regression algorithm from Spark ML .

Information at a Glance

Note: This operator can only be used with TIBCO® Data Virtualization and Apache Spark 3.2 or later.

Parameter	Description
Category	Model
Data source type	TIBCO® Data Virtualization
Send output to other operators	Yes
Data processing tool	TIBCO® DV, Apache Spark 3.2 or later

Algorithm

The Gradient-Boosted Tree algorithm is a predictive method, by which a series of shallow regression trees incrementally reduces prediction errors of previous trees. This operator implements the open-source Gradient-Boosted Tree Regression (GBTR) algorithm from Spark MLlib.

Input

An input is a single tabular data set.

Bad or Missing Values

Null values are not allowed and result in an error.
The Max Bins parameter should be increased to the maximum cardinality of categorical features. However, depending on the available resources, the system might not be able to handle very high values results in an error.
If the number of levels in the dependent column is not equal to 2, an error is reported.

Configuration

Users can fine-tune the hyper-parameters of interest with the cross-validation training method and utilize the chosen metric to evaluate the performance. The following table includes the configuration details for the Gradient-Boosted Tree Regression operator.

Parameter	Description
Notes	Notes or helpful information about this operator's parameter settings. When you enter content in the Notes field, a yellow asterisk appears on the operator.
Dependent Variable	Specify the categorical data column as a dependent column.
Use all available columns as Predictors	When set to Yes, the operator uses all the available columns as predictors and ignores the Continuous Predictors and Categorical Predictors parameters. When set to No, the user must select at least one of the Continuous or Categorical Predictors.
Continuous Predictors	Specify the numerical data columns as independent columns. It must be numerical column. Click Select Columns to select the required columns. Note: The columns selected in the Categorical Predictors parameter are not available.
Categorical Predictors	Specify the categorical data columns as independent columns. Note: The columns selected in the Continuous Predictors parameter are not available.
Evaluation Metric	The metric for evaluating regression models. The following values are available: MAE MSE R2 RMSE Default: RMSE
Loss Function	The loss function to minimize. The following values are available: L1 (Absolute-value loss function L2 (Squared-value loss function Default: L1
Number of Trees	A string specifying the number of trees. The input for this parameter should be a comma-separated sequence of integer values (for example, 10, 100.) Default: 100
Number of Feature Functions	A function to determine the number of features for building each decision tree. The following values are available: All ⅓ Square Root Log2 User Defined Default: Square Root
Feature Sampling Ratio	The fraction of the number of features per node to use when the Number of Feature Functions is set to User Defined. The input for this parameter should be a comma-separated sequence of double values in (0,1). Default: 0.5, 0.7
Max Depth	The maximum depth of each tree. The input for this parameter should be a comma-separated sequence of integer values. Default: 3, 5
Row Sampling Ratio	The fraction of training data for building each decision tree. The input for this parameter should be a comma-separated sequence of double values in (0,1). Default: 1
Min Leaf Size	The smallest number of data instances that can exist within a terminal leaf node of a decision tree. The input for this parameter should be a comma-separated sequence of integer values (for example, 1,2). Default: 1
Max Bins	The maximum number of bins used for discretizing and splitting continuous features. The input for this parameter should be a comma-separated sequence of integer values (for example, 256). The number of Max Bins should be larger than the number of unique levels of any selected categorical columns. Default: 32
Learning Rate	The shrinkage parameter to control the contribution of each estimator. The input for this parameter should be a comma-separated sequence of double values in the interval (0,1). Default: 0.1
Number of Cross Validation Folds	The number of cross-validation samples. Default: 3
Random Seed	The seed used for the pseudo-random row extraction. Default: 1

Output

Visual Output

Parameter Summary Info Displays information about the input parameters and their current settings.
Variable Importance Displays the importance of predictors as evaluated in the training process. For each predictor, the importance of the model is shown in the second column.
Training Summary Displays a table with a row for each tested combination of hyper-parameters. For each hyper-parameter, the chosen metric is displayed and the Best Model is marked.

Output to successive operators

The output of the operator is the model object with the best validation performance. You can use this regression model object with a compatible Predictor operator. One additional column, PRED_GBTR (the predictive value of the regression model), is produced in the Predictor operator.

A regression model object that can also be used with a Regression Evaluator operator.