The ML.FEATURE_INFO function

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

Syntax

ML.FEATURE_INFO(MODEL `project_id.dataset.model`)

Arguments

ML.FEATURE_INFO takes the following arguments:

  • project_id: Your project ID.
  • dataset: The BigQuery dataset that contains the model.
  • model: The name of the model.

Output

ML.FEATURE_INFO returns the following columns:

  • input: a STRING value that contains the name of the column in the input training data.
  • min: a FLOAT64 value that contains the minimum value in the input column. min is NULL for non-numeric inputs.
  • max: a FLOAT64 value that contains the maximum value in the input column. max is NULL for non-numeric inputs.
  • mean: a FLOAT64 value that contains the average value for the input column. mean is NULL for non-numeric inputs.
  • median: a FLOAT64 value that contains the median value for the input column. median is NULL for non-numeric inputs.
  • stddev: a FLOAT64 value that contains the standard deviation value for the input column. stddev is NULL for non-numeric inputs.
  • category_count: an INT64 value that contains the number of categories in the input column. category_count is NULL for non-categorical columns.
  • null_count: an INT64 value that contains the number of NULL values in the input column.
  • dimension: an INT64 value that contains the dimension of the input column if the input column has a ARRAY<STRUCT> type. dimension is NULL for non-ARRAY<STRUCT> columns.

For matrix factorization models, only category_count is calculated for the user and item columns.

If you used the TRANSFORM clause in the CREATE MODEL statement that created the model, ML.FEATURE_INFO outputs the information of the pre-transform columns from the query_statement argument.

Permissions

You must have the bigquery.models.create and bigquery.models.getData Identity and Access Management (IAM) permissions in order to run ML.FEATURE_INFO.

Limitations

ML.FEATURE_INFO doesn't support imported TensorFlow models.

Example

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

SELECT
  *
FROM
  ML.FEATURE_INFO(MODEL `mydataset.mymodel`)

What's next