Categories:: Model monitor functions

MODEL_MONITOR_DRIFT_METRIC¶

Gets drift metrics from a model monitor. Each model monitor monitors one machine learning model.

See also:: Querying monitoring results for more information.

Syntax¶

MODEL_MONITOR_DRIFT_METRIC(
  <model_monitor_name>, <drift_metric_name>, <column_name>
  [ , <granularity> [ , <start_time>  [ , <end_time> ] ] ]
)

Copy

Arguments¶

Required:

model_monitor_name

Name of the model monitor used to compute the metric.

Valid values: A string that’s the name of the model monitor. It can be a simple or fully qualified name.

drift_metric_name

Name of the metric.

Valid values:

'JENSEN_SHANNON'
'DIFFERENCE_OF_MEANS'
'WASSERSTEIN'

column_name

Name of the column used to compute drift.

Valid values: Any string that exists as a feature column, prediction column, or actual column in the model monitor.

Optional:

granularity

Granularity of the time range being queried.

Valid values:

'<num> DAY'
'<num> WEEK'
'<num> MONTH'
'<num> QUARTER'
'<num> YEAR'
'ALL'
NULL

start_time

Start of the time range used to compute the metric.

Valid values: A constant timestream expression or NULL.

end_time

End of the time range used to compute the metric.

Valid values: A constant timestream expression or NULL.

Returns¶

Column	Description	Example values
`EVENT_TIMESTAMP`	Timestamp at the start of the time range.	`2024-01-01 00:00:00.000`
`METRIC_VALUE`	Value of the metric within the specified time range.	`5`
`COL_COUNT_USED`	Number of records used to compute the metric.	`100`
`COL_COUNT_UNUSED`	Number of records excluded from the metric computation.	`10`
`BASELINE_COL_COUNT_USED`	Number of records used to compute the metric.	`10`
`BASELINE_COL_COUNT_UNUSED`	Number of records excluded from the metric computation.	`0`
`METRIC_NAME`	Name of the drift metric that has been computed.	`DIFFERENCE_OF_MEANS`
`COLUMN_NAME`	Name of the column for which the drift metric has been computed.	`FEATURE_NAME`

Usage Notes¶

The model monitor must have a baseline set for the drift metric to be computed.

You might run into errors if you:

Don’t set a baseline for the model monitor.
Requested a numerical drift metric for a non-numeric feature.
Use a drift metric that doesn’t exist in the model monitor.

Examples¶

The following example gets the differences of means drift metric for MY_MONITOR over a one-day period:

SELECT * FROM TABLE(MODEL_MONITOR_DRIFT_METRIC(
'MY_MONITOR', 'DIFFERENCE_OF_MEANS', 'MODEL_PREDICTION', '1 DAY', TO_TIMESTAMP_TZ('2024-01-01'), TO_TIMESTAMP_TZ('2024-01-02'))
)

Copy