LIGHTGRADBOOST Procedure

AUTOTUNE Statement

AUTOTUNE <options>;

The AUTOTUNE statement searches for the best combination of values of the BAGGINGFRACTION=, BAGGINGFREQUENCY=, INPUTFRACTION=, LASSO=, LEAFSIZE=, LEARNINGRATE=, MAXDEPTH=, MAXITERS=, NUMBIN=, and RIDGE= options in the PROC LIGHTGRADBOOST statement. These options control the training algorithm and are often called hyperparameters to distinguish them from the parameters of the trained model. Optimal hyperparameter settings depend on the data and are never known for certain.

Table 3 summarizes the options that you can specify in the AUTOTUNE statement. For more information about all options except the TUNINGPARAMETERS= option, see the option’s description in the section AUTOTUNE Statement in Chapter 2, Shared Concepts. The TUNINGPARAMETERS= option is described in Table 3.

Table 3: AUTOTUNE Statement Options

Option	Description
APPENDLOOKUP	Specifies that the table specified in the HISTORYTABLE= option contain the rows from the table specified in the LOOKUPTABLE= option
CONSTRAINTS=	Specifies the constraints on the output metrics
EVALHISTORY=	Specifies how to report the evaluation history of the tuner
FOLDCOLUMN=	Specifies the column in the data table in which the cross validation fold value is indicated
FRACTION=	Specifies the fraction of observations to use for validation
HISTORYTABLE=	Specifies the table that contains the evaluation history
KFOLD=	Specifies the number of folds for k-fold cross validation
LIVEUPDATE	Specifies that the table specified in the HISTORYTABLE= option be updated at every evaluation
LOOKUPTABLE=	Specifies the table to use for evaluation lookup
MAXBAYES=	Specifies the maximum number of points in the kriging model
MAXEVALS=	Specifies the maximum number of evaluations
MAXITER=	Specifies the maximum number of iterations when SEARCHMETHOD=GA or SEARCHMETHOD=BAYESIAN
MAXTIME=	Specifies the maximum time for all iterations
MAXTRAINTIME=	Specifies the maximum time for a model training
NCONVITER=	Specifies the number of convergence iterations
NOGRIDSHUFFLE	Requests that the grid points not be shuffled
NOLOCALSEARCH	Disables local search optimization
NPARALLEL=	Specifies the number of parallel sessions
NSUBSESSIONWORKERS=	Specifies the number of workers in parallel sessions
OBJECTIVE=	Specifies the objective function
POPSIZE=	Specifies the population size when SEARCHMETHOD=GA or SEARCHMETHOD=BAYESIAN
SAMPLESIZE=	Specifies the sample size when SEARCHMETHOD=LHS or SEARCHMETHOD=RANDOM
SEARCHMETHOD=	Specifies the search method that the optimizer uses
SECONDOBJECTIVE=	Specifies the second objective to use for tuning
SELECTINITPOINT	Specifies that the tuner select the best evaluation from the lookup table
TARGETEVENT=	Specifies the target event for ROC-based calculations
TRAINFRACTION=	Specifies the fraction of observations to use for training
TUNINGPARAMETERS=	Specifies the custom tuning parameters
USEPARAMETERS=	Specifies how to handle the TUNINGPARAMETERS= option

TUNINGPARAMETERS=(suboption |…|<suboption>) TUNEPARMS=(suboption |…|<suboption>)

specifies which parameters to tune and which ranges to tune over. If you specify USEPARAMETERS=STANDARD, this option is ignored.

You can specify one or more of the following suboptions:

BAGGINGFRACTION (LB=number UB=number VALUES=value-list INIT=number EXCLUDE)

specifies information about the BAGGINGFRACTION= option values to use in tuning the light gradient boosting model. For more information, see the BAGGINGFRACTION= option in the PROC LIGHTGRADBOOST statement.

You can specify the following additional suboptions:

LB=number

specifies the minimum BAGGINGFRACTION= option value to consider during tuning. If you specify this suboption, you cannot specify the VALUES= suboption.

By default, LB=0.2

UB=number

specifies the maximum BAGGINGFRACTION= option value to consider during tuning. If you specify this suboption, you cannot specify the VALUES= suboption.

By default, UB=0.95

VALUES=value-list

specifies a list of BAGGINGFRACTION= option values to consider during tuning, where value-list is a space-separated list of numbers greater than or equal to 0 and less than or equal to 1. If you specify this suboption, you cannot specify either the LB= or UB= suboption.

INIT=number

specifies the initial BAGGINGFRACTION= option value for the tuner to use.

By default, INIT=0.95

EXCLUDE

excludes the BAGGINGFRACTION= option from the tuning process. If you specify this suboption, any specified LB=, UB=, VALUES=, and INIT= suboptions are ignored.

BAGGINGFREQUENCY (LB=number UB=number VALUES=value-list INIT=number EXCLUDE)

specifies information about the BAGGINGFREQUENCY= option values to use in tuning the light gradient boosting model. For more information, see the BAGGINGFREQUENCY= option in the PROC LIGHTGRADBOOST statement.

You can specify the following additional suboptions:

LB=number

specifies the minimum BAGGINGFREQUENCY= option value to consider during tuning. If you specify this suboption, you cannot specify the VALUES= suboption.

By default, LB=0

UB=number

specifies the maximum BAGGINGFREQUENCY= option value to consider during tuning. If you specify this suboption, you cannot specify the VALUES= suboption.

By default, UB=7

VALUES=value-list

specifies a list of BAGGINGFREQUENCY= option values to consider during tuning, where value-list is a space-separated list of numbers greater than or equal to 0. If you specify this suboption, you cannot specify either the LB= or UB= suboption.

INIT=number

specifies the initial BAGGINGFREQUENCY= option value for the tuner to use.

By default, INIT=0

EXCLUDE

excludes the BAGGINGFREQUENCY= option from the tuning process. If you specify this suboption, any specified LB=, UB=, VALUES=, and INIT= suboptions are ignored.

INPUTFRACTION (LB=number UB=number VALUES=value-list INIT=number EXCLUDE)

specifies information about the INPUTFRACTION= option values to use in tuning the light gradient boosting model. For more information, see the INPUTFRACTION= option in the PROC LIGHTGRADBOOST statement.

You can specify the following additional suboptions:

LB=number

specifies the minimum INPUTFRACTION= option value to consider during tuning. If you specify this suboption, you cannot specify the VALUES= suboption.

By default, LB=0.05

UB=number

specifies the maximum INPUTFRACTION= option value to consider during tuning. If you specify this suboption, you cannot specify the VALUES= suboption.

By default, UB=1.0

VALUES=value-list

specifies a list of INPUTFRACTION= option values to consider during tuning, where value-list is a space-separated list of numbers greater than 0 and less than or equal to 1. If you specify this suboption, you cannot specify either the LB= or UB= suboption.

INIT=number

specifies the initial INPUTFRACTION= option value for the tuner to use.

By default, INIT=1.0

EXCLUDE

excludes the INPUTFRACTION= option from the tuning process. If you specify this suboption, any specified LB=, UB=, VALUES=, and INIT= suboptions are ignored.

LASSO (LB=number UB=number VALUES=value-list INIT=number EXCLUDE)

specifies information about the LASSO= option values to use in tuning the light gradient boosting model. For more information, see the LASSO= option in the PROC LIGHTGRADBOOST statement.

You can specify the following additional suboptions:

LB=number

specifies the minimum LASSO= option value to consider during tuning. If you specify this suboption, you cannot specify the VALUES= suboption.

By default, LB=0.

UB=number

specifies the maximum LASSO= option value to consider during tuning. If you specify this suboption, you cannot specify the VALUES= suboption.

By default, UB=10.

VALUES=value-list

specifies a list of LASSO= option values to consider during tuning, where value-list is a space-separated list of numbers greater than or equal to 0. If you specify this suboption, you cannot specify either the LB= or UB= suboption.

INIT=number

specifies the initial LASSO= option value for the tuner to use.

By default, INIT=0.

EXCLUDE

excludes the LASSO= option from the tuning process. If you specify this suboption, any specified LB=, UB=, VALUES=, and INIT= suboptions are ignored.

LEAFSIZE (LB=number UB=number VALUES=value-list INIT=number EXCLUDE)

specifies information about the LEAFSIZE= option values to use in tuning the light gradient boosting model. This option is not tuned by default. If it is specified without any suboptions, then the tuner uses the following values: 1, 5, 10, 20, 40, 80, 160, and 320. For more information, see the LEAFSIZE= option in the PROC LIGHTGRADBOOST statement.

You can specify the following additional suboptions:

LB=number

specifies the minimum LEAFSIZE= option value to consider during tuning. If you specify this suboption, you cannot specify the VALUES= suboption.

By default, LB=1.

UB=number

specifies the maximum LEAFSIZE= option value to consider during tuning. If you specify this suboption, you cannot specify the VALUES= suboption.

By default, UB=320.

VALUES=value-list

specifies a list of LEAFSIZE= option values to consider during tuning, where value-list is a space-separated list of integers greater than or equal to 0. If you specify this suboption, you cannot specify either the LB= or UB= suboption.

INIT=number

specifies the initial LEAFSIZE= option for the tuner to use.

By default, INIT=20.

EXCLUDE

excludes the LEAFSIZE= option from the tuning process.

LEARNINGRATE (LB=number UB=number VALUES=value-list INIT=number EXCLUDE)

specifies information about the LEARNINGRATE= option values to use in tuning the light gradient boosting model. For more information, see the LEARNINGRATE= option in the PROC LIGHTGRADBOOST statement.

You can specify the following additional suboptions:

LB=number

specifies the minimum LEARNINGRATE= option value to consider during tuning. If you specify this suboption, you cannot specify the VALUES= suboption.

By default, LB=0.01.

UB=number

specifies the maximum LEARNINGRATE= option value to consider during tuning. If you specify this suboption, you cannot specify the VALUES= suboption.

By default, UB=1.

VALUES=value-list

specifies a list of LEARNINGRATE= option values to consider during tuning, where value-list is a space-separated list of numbers greater than 0 and less than or equal to 1. If you specify this suboption, you cannot specify either the LB= or UB= suboption.

INIT=number

specifies the initial LEARNINGRATE= option value for the tuner to use.

By default, INIT=0.1.

EXCLUDE

excludes the LEARNINGRATE= option from the tuning process. If you specify this suboption, any specified LB=, UB=, VALUES=, and INIT= suboptions are ignored.

MAXDEPTH (LB=number UB=number VALUES=value-list INIT=number EXCLUDE)

specifies information about the MAXDEPTH= option value to use in tuning the light gradient boosting model. If it is specified without any suboptions, then the tuner uses the following values: –1, 3, 4, 5, 6, and 7. For more information, see the MAXDEPTH= option in the PROC LIGHTGRADBOOST statement.

You can specify the following additional suboptions:

LB=number

specifies the minimum MAXDEPTH= option value to consider during tuning. If you specify this suboption, you cannot specify the VALUES= suboption.

By default, LB=–1.

UB=number

specifies the maximum MAXDEPTH= option value to consider during tuning. If you specify this suboption, you cannot specify the VALUES= suboption.

By default, UB=7.

VALUES=value-list

specifies a list of MAXDEPTH= option values to consider during tuning, where value-list is a space-separated list of integers greater than or equal to 1. If you specify this suboption, you cannot specify either the LB= or UB= suboption.

INIT=number

specifies the initial MAXDEPTH= option value for the tuner to use.

By default, INIT=–1.

EXCLUDE

excludes the MAXDEPTH= option from the tuning process. If you specify this suboption, any specified LB=, UB=, VALUES=, and INIT= suboptions are ignored.

NUMBIN (LB=number UB=number VALUES=value-list INIT=number EXCLUDE)

specifies information about the NUMBIN= option values to use in tuning the light gradient boosting model. For more information, see the NUMBIN= option in the PROC LIGHTGRADBOOST statement.

You can specify the following additional suboptions:

LB=number

specifies the minimum NUMBIN= option value to consider during tuning. If you specify this suboption, you cannot specify the VALUES= suboption.

By default, LB=2.

UB=number

specifies the maximum NUMBIN= option value to consider during tuning. If you specify this suboption, you cannot specify the VALUES= suboption.

By default, UB=255.

VALUES=value-list

specifies a list of the NUMBIN= option values to consider during tuning, where value-list is a space-separated list of positive integers. If you specify this suboption, you cannot specify either the LB= or UB= suboption.

INIT=number

specifies the initial NUMBIN= option value for the tuner to use.

By default, INIT=255.

EXCLUDE

excludes the NUMBIN= option from the tuning process. If you specify this suboption, any specified LB=, UB=, VALUES=, and INIT= suboptions are ignored.

RIDGE (LB=number UB=number VALUES=value-list INIT=number EXCLUDE)

specifies information about the RIDGE= option values to use in tuning the light gradient boosting model. For more information, see the RIDGE= option in the PROC LIGHTGRADBOOST statement.

You can specify the following additional suboptions:

LB=number

specifies the minimum RIDGE= option value to consider during tuning. If you specify this suboption, you cannot specify the VALUES= suboption.

By default, LB=0.

UB=number

specifies the maximum RIDGE= option value to consider during tuning. If you specify this suboption, you cannot specify the VALUES= suboption.

By default, UB=10.

VALUES=value-list

specifies a list of RIDGE= option values to consider during tuning, where value-list is a space-separated list of numbers greater than or equal to 0. If you specify this suboption, you cannot specify either the LB= or UB= suboption.

INIT=number

specifies the initial RIDGE= option value for the tuner to use.

By default, INIT=0.

EXCLUDE

excludes the RIDGE= option from the tuning process. If you specify this suboption, any specified LB=, UB=, VALUES=, and INIT= suboptions are ignored.

Last updated: September 23, 2026