The ML.ROBUST_SCALER function

This document describes theML.ROBUST_SCALER function, which lets you scale anumerical expression by using statistics that are robust to outliers. Thefunction performs the scaling by removing themedian and scalingthe data according to thequantilerange.

When used in theTRANSFORM clause,the median and quantile range calculated during training are automaticallyused in prediction.

You can use this function with models that supportmanual feature preprocessing. For moreinformation, see the following documents:

• End-to-end user journeys for ML models
• Contribution analysis user journey

Syntax

ML.ROBUST_SCALER(numerical_expression [, quantile_range] [, with_median] [, with_quantile_range]) OVER()

Arguments

ML.ROBUST_SCALER takes the following arguments:

• numerical_expression: thenumericalexpression to scale.
• quantile_range: an array of twoINT64 elements that specifiesthe quantile range. The first element provides the lower boundary of therange. It must be greater than0. The second element provides the upperboundary of the range. It must be greater than the first element but lessthan100. The default value is[25, 75].
• with_median: aBOOL value that specifies whether the datais centered. IfTRUE, the function centers the data by removing themedian before scaling. The default value isTRUE.
• with_quantile_range: aBOOL value that specifies whether thedata is scaled to the quantile range. IfTRUE, the data is scaled. Thedefault value isTRUE.

Output

ML.ROBUST_SCALER returns aFLOAT64 value that represents the scalednumerical expression.

Example

The following example centers a set of numerical expressions and thenscales it to the range[25, 75]:

SELECTf,ML.ROBUST_SCALER(f)OVER()ASoutputFROMUNNEST([NULL,-3,1,2,3,4,5])ASfORDERBYf;

The output looks similar to the following:

+------+---------------------+|  f   |       output        |+------+---------------------+| NULL |                NULL ||   -3 | -1.6666666666666667 ||    1 | -0.3333333333333333 ||    2 |                 0.0 ||    3 |  0.3333333333333333 ||    4 |  0.6666666666666666 ||    5 |                 1.0 |+------+---------------------+

What's next

• For information about feature preprocessing, seeFeature preprocessing overview.