The ML.FEATURE_INFO function

This document describes theML.FEATURE_INFO function, which lets you seeinformation about the input features that are used to train a model.

For more information about which models support this function, seeEnd-to-end user journeys for ML models.

Syntax

ML.FEATURE_INFO(MODEL `PROJECT_ID.DATASET.MODEL_NAME`)

Arguments

ML.FEATURE_INFO takes the following arguments:

• PROJECT_ID: Your project ID.
• DATASET: The BigQuery datasetthat contains the model.
• MODEL_NAME: The name of the model.

Output

ML.FEATURE_INFO returns the following columns:

• input: aSTRING value that contains the name of the column in theinput training data.
• min: aFLOAT64 value that contains the minimum value in theinputcolumn.min isNULL for non-numeric inputs.
• max: aFLOAT64 value that contains the maximum value in theinputcolumn.max isNULL for non-numeric inputs.
• mean: aFLOAT64 value that contains the average value for theinputcolumn.mean isNULL for non-numeric inputs.
• median: aFLOAT64 value that contains the median value for theinputcolumn.median isNULL for non-numeric inputs.
• stddev: aFLOAT64 value that contains the standard deviation value fortheinput column.stddev isNULL for non-numeric inputs.
• category_count: anINT64 value that contains the number of categoriesin theinput column.category_count isNULL for non-categorical columns.
• null_count: anINT64 value that contains the number ofNULL valuesin theinput column.
• dimension: anINT64 value that contains the dimension of theinputcolumn if theinput column has aARRAY<STRUCT> type.dimension isNULL for non-ARRAY<STRUCT> columns.

Formatrix factorizationmodels, onlycategory_count is calculated for theuser anditemcolumns.

If you used theTRANSFORM clausein theCREATE MODEL statement that created the model,ML.FEATURE_INFOoutputs the information of the pre-transform columns from thequery_statement argument.

Permissions

You must have thebigquery.models.create andbigquery.models.getDataIdentity and Access Management (IAM) permissionsin order to runML.FEATURE_INFO.

Limitations

ML.FEATURE_INFO doesn't supportimported TensorFlow models.

Example

The following example retrieves feature information from the modelmydataset.mymodel in your default project:

SELECT*FROMML.FEATURE_INFO(MODEL`mydataset.mymodel`)

What's next

• For information about feature preprocessing, seeFeature preprocessing overview.