The ML.FEATURE_IMPORTANCE function

This document describes theML.FEATURE_IMPORTANCE function, which lets yousee the feature importance score. This score indicates how useful or valuableeach feature was in the construction of a boosted tree or a random forest modelduring training. For more information, see thefeature_importances propertyin the XGBoost library.

Syntax

ML.FEATURE_IMPORTANCE(  MODEL `PROJECT_ID.DATASET.MODEL`)

Arguments

ML.FEATURE_IMPORTANCE takes the following arguments:

• PROJECT_ID: your project ID.
• DATASET: the BigQuery dataset that containsthe model.
• MODEL: the name of the model.

Output

ML.FEATURE_IMPORTANCE returns the following columns:

• feature: aSTRING value that contains the name of the feature columnin the input training data.
• importance_weight: aFLOAT64 value that contains the number of times afeature is used to split the data across all trees.
• importance_gain: aFLOAT64 value that contains the average gain across allsplits the feature is used in.
• importance_cover: aFLOAT64 value that contains the average coverageacross all splits the feature is used in.

If theTRANSFORM clausewas used in theCREATE MODEL statement that created the model,ML.FEATURE_IMPORTANCE returns the information of the pre-transform columnsfrom thequery_statement clause of theCREATE MODEL statement.

Permissions

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

Limitations

ML.FEATURE_IMPORTANCE is only supported withboosted tree modelsandrandom forest models.

Example

This example retrieves feature importance frommymodel inmydataset. The dataset is in your default project.

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

What's next

• For more information about Explainable AI, seeBigQuery Explainable AI overview.
• For more information about supported SQL statements and functions for MLmodels, seeEnd-to-end user journeys for ML models.