Monitor features

Vertex AI Feature Store lets you schedule and run feature monitoring jobs to monitor feature data, retrieve feature statistics, and detect feature drift. You can monitor feature data only if you've registered your feature data source in the Feature Registry.

To monitor feature data, you can create the FeatureMonitor resource under a FeatureGroup resource. While creating the FeatureMonitor resource, you can configure the monitoring schedule to periodically run monitoring jobs on the feature data. Alternatively, you can run a feature monitoring job manually to monitor your feature data outside of the monitoring schedule.

For each monitoring job that's executed, Vertex AI Feature Store generates a FeatureMonitorJob resource, which you can retrieve to view the feature statistics and information about drift detected in the feature data.

Before you begin

Before you monitor features using Vertex AI Feature Store, complete the prerequisites listed in this section.

Register feature data source

Register your feature data source from BigQuery in the Feature Registry by creating feature groups and and features. The FeatureMonitor resources used to retrieve and monitor feature statistics are associated with feature groups.

Authenticate to Vertex AI

Authenticate to Vertex AI, unless you've done so already.

Select the tab for how you plan to use the samples on this page:

Python

To use the Python samples on this page in a local development environment, install and initialize the gcloud CLI, and then set up Application Default Credentials with your user credentials.

  1. Install the Google Cloud CLI.

  2. To initialize the gcloud CLI, run the following command: 

    gcloud init

  3. If you're using a local shell, then create local authentication credentials for your user account: 

    gcloud auth application-default login

    You don't need to do this if you're using Cloud Shell.

For more information, see Set up authentication for a local development environment.

REST

To use the REST API samples on this page in a local development environment, you use the credentials you provide to the gcloud CLI.

    Install the Google Cloud CLI, then initialize it by running the following command: 

    gcloud init

For more information, see Authenticate for using REST in the Google Cloud authentication documentation.

Create a feature monitor with a monitoring schedule

To retrieve and monitor feature statistics, create a FeatureMonitor resource specifying the schedule to periodically execute feature monitoring jobs and retrieve feature statistics for the features registered in the feature group.

Use the following samples to create a FeatureMonitor resource. To set up multiple schedules for the same feature group, you must create multiple FeatureMonitor resources.

REST

To create a FeatureMonitor resource and schedule feature monitoring jobs, send a POST request by using the featureMonitors.create method.

Before using any of the request data, make the following replacements:

  • LOCATION_ID: Region where you want to create the feature monitor, such as us-central1.
  • PROJECT_ID: Your project ID.
  • FEATUREGROUP_NAME: The name of the feature group where you set up feature monitoring.
  • FEATURE_MONITOR_NAME: A name for the new feature monitor that you want to create.
  • FEATURE_ID_1 and FEATURE_ID_2: The IDs of the features that you want to monitor.
  • DRIFT_THRESHOLD_1 and DRIFT_THRESHOLD_2: Drift thresholds for each feature included in the feature monitor. The drift threshold is used to detect anomalies, such as feature drift. Enter a value in the range [0, 1). If you don't enter a value, the threshold is set to 0.3, by default.
    Vertex AI Feature Store compares the snapshots from consecutive feature monitor job executions and calculates drifts using the ML.TFDV_VALIDATE function in BigQuery. To classify anomalies, L-infinity distance is used for categorical features and Jensen-Shannon divergence is used for numerical features.
  • CRON: Cron schedule expression representing the frquency for running the feature monitoring job. For more information, see cron.

HTTP method and URL: 

POST https://LOCATION_ID-aiplatform.googleapis.com/v1beta1/projects/PROJECT_ID/locations/LOCATION_ID/featureGroups/FEATUREGROUP_NAME/featureMonitors?feature_monitor_id=FEATURE_MONITOR_NAME

Request JSON body: 

{
  "feature_selection_config": {
    "feature_configs": [
      {"feature_id":"FEATURE_ID_1", "drift_threshold": "DRIFT_THRESHOLD_1" },
      {"feature_id":"FEATURE_ID_2", "drift_threshold": "DRIFT_THRESHOLD_2" }
    ],
  },
  "schedule_config": {
    "cron": "CRON"
  }
}

To send your request, choose one of these options:

curl

Save the request body in a file named request.json, and execute the following command: 

curl -X POST \
     -H "Authorization: Bearer $(gcloud auth print-access-token)" \
     -H "Content-Type: application/json; charset=utf-8" \
     -d @request.json \
     "https://LOCATION_ID-aiplatform.googleapis.com/v1beta1/projects/PROJECT_ID/locations/LOCATION_ID/featureGroups/FEATUREGROUP_NAME/featureMonitors?feature_monitor_id=FEATURE_MONITOR_NAME"

PowerShell

Save the request body in a file named request.json, and execute the following command: 

$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred" }

Invoke-WebRequest `
    -Method POST `
    -Headers $headers `
    -ContentType: "application/json; charset=utf-8" `
    -InFile request.json `
    -Uri "https://LOCATION_ID-aiplatform.googleapis.com/v1beta1/projects/PROJECT_ID/locations/LOCATION_ID/featureGroups/FEATUREGROUP_NAME/featureMonitors?feature_monitor_id=FEATURE_MONITOR_NAME" | Select-Object -Expand Content

You should receive a JSON response similar to the following:

{
  "name": "projects/PROJECT_NUMBER/locations/LOCATION_ID/featureGroups/FEATUREGROUP_NAME/featureMonitors/FEATURE_MONITOR_NAME/operations/OPERATION_ID",
  "metadata": {
    "@type": "type.googleapis.com/google.cloud.aiplatform.v1beta1.CreateFeatureMonitorOperationMetadata",
    "genericMetadata": {
      "createTime": "2024-12-15T19:35:03.975958Z",
      "updateTime": "2024-12-15T19:35:03.975958Z"
    }
  }
}

Python

Before trying this sample, follow the Python setup instructions in the Vertex AI quickstart using client libraries. For more information, see the Vertex AI Python API reference documentation.

To authenticate to Vertex AI, set up Application Default Credentials. For more information, see Set up authentication for a local development environment. 

from google.cloud import aiplatform
from vertexai.resources.preview import feature_store

def create_feature_monitor_sample(
    project: str,
    location: str,
    existing_feature_group_id: str,
    feature_monitor_id: str,
    feature_selection_configs: List[Tuple[str, float]]
    schedule_config: str # Cron string. For example, "0 * * * *" indicates hourly execution.
):
    aiplatform.init(project="PROJECT_ID", location="LOCATION_ID")
    feature_group = feature_store.FeatureGroup("FEATUREGROUP_NAME")
    feature_monitor = feature_group.create_feature_monitor(
        name= "FEATURE_MONITOR_NAME",
        feature_selection_configs=[("FEATURE_ID_1", DRIFT_THRESHOLD_1),("FEATURE_ID_2", DRIFT_THRESHOLD_2)],
        schedule_config="CRON"
        )

Replace the following:

  • LOCATION_ID: Region where you want to create the feature monitor, such as us-central1.
  • PROJECT_ID: Your project ID.
  • FEATUREGROUP_NAME: The name of the feature group where you set up feature monitoring.
  • FEATURE_MONITOR_NAME: A name for the new feature monitor that you want to create.
  • FEATURE_ID_1 and FEATURE_ID_2: The IDs of the features that you want to monitor.
  • DRIFT_THRESHOLD_1 and DRIFT_THRESHOLD_2: Drift thresholds for each Feature included in the feature monitor. The drift threshold is used to detect feature drift. Enter a value between 0 and 1. If you don't enter a value, the threshold is set to 0.3, by default.
    Vertex AI Feature Store compares the data snapshot from the current feature monitor job with the data snapshot during the previous feature monitor job. Note that to calculate the drift score, Vertex AI Feature Store uses the ML.TFDV_VALIDATE function in BigQuery.
    For the metric used to compare statistics, L-infinity distance is used for categorical features and Jensen-Shannon divergence is used for numerical features.
  • CRON: Cron schedule expression representing the frequency for running the feature monitoring job. For more information, see cron.

Run a feature monitoring job manually

You can skip the wait between consecutive scheduled feature monitoring jobs and manually run a feature monitor job. This is useful if you want to retrieve monitoring information and detect anomalies in the feature data immediately instead of waiting for the next scheduled monitoring job to run.

REST

To run a feature monitoring job manually by creating a FeatureMonitorJob resource , send a POST request by using the featureMonitorJobs.create method.

Before using any of the request data, make the following replacements:

  • LOCATION_ID: Region where you want to run the feature monitoring job, such as us-central1.
  • FEATUREGROUP_NAME: The name of the feature group containing the FeatureMonitor resource.
  • PROJECT_ID: Your project ID.
  • FEATURE_MONITOR_NAME: The name of the FeatureMonitor resource for which you want to run the feature monitoring job.

HTTP method and URL: 

POST https://LOCATION_ID-aiplatform.googleapis.com/v1beta1/projects/PROJECT_ID/locations/LOCATION_ID/featureGroups/FEATUREGROUP_NAME/featureMonitors/FEATURE_MONITOR_ID/featureMonitorJobs

To send your request, choose one of these options:

curl

Execute the following command: 

curl -X POST \
     -H "Authorization: Bearer $(gcloud auth print-access-token)" \
     -H "Content-Type: application/json; charset=utf-8" \
     -d "" \
     "https://LOCATION_ID-aiplatform.googleapis.com/v1beta1/projects/PROJECT_ID/locations/LOCATION_ID/featureGroups/FEATUREGROUP_NAME/featureMonitors/FEATURE_MONITOR_ID/featureMonitorJobs"

PowerShell

Execute the following command: 

$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred" }

Invoke-WebRequest `
    -Method POST `
    -Headers $headers `
    -Uri "https://LOCATION_ID-aiplatform.googleapis.com/v1beta1/projects/PROJECT_ID/locations/LOCATION_ID/featureGroups/FEATUREGROUP_NAME/featureMonitors/FEATURE_MONITOR_ID/featureMonitorJobs" | Select-Object -Expand Content

You should receive a JSON response similar to the following:

{
  "name": "projects/PROJECT_NUMBER/locations/LOCATION_ID/featureGroups/FEATUREGROUP_NAME/featureMonitors/FEATURE_MONITOR_NAME/featureMonitorJobs/FEATURE_MONITOR_JOB_ID"
}

Python

Before trying this sample, follow the Python setup instructions in the Vertex AI quickstart using client libraries. For more information, see the Vertex AI Python API reference documentation.

To authenticate to Vertex AI, set up Application Default Credentials. For more information, see Set up authentication for a local development environment. 

from google.cloud import aiplatform
from vertexai.resources.preview import feature_store

aiplatofrm.init(project="PROJECT_ID", location="LOCATION_ID")

feature_group = FeatureGroup.get("FEATUREGROUP_NAME}")
feature_monitor = feature_group.get_feature_monitor(FEATURE_MONITOR_NAME)
feature_monitor_job = feature_monitor.create_feature_monitor_job()

Replace the following:

  • LOCATION_ID: Region where you want to run the feature monitoring job, such as us-central1.
  • PROJECT_ID: Your project ID.
  • FEATUREGROUP_NAME: The name of the feature group containing the FeatureMonitor resource.
  • FEATURE_MONITOR_NAME: The name of the FeatureMonitor resource for which you want to run the feature monitoring job.

Retrieve feature statistics from a monitoring job

You can retrieve feature statistics for all the features in a feature monitoring job by retrieving the FeatureMonitorJob resource using the feature monitor job ID generated during the feature monitoring job execution. You can also retrieve feature statistics for a specific resource for the latest monitoring job.

List feature monitor jobs

The following samples show how to retrieve a list of all the FeatureMonitorJob resources created for a given FeatureMonitor resource.

REST

To retrieve a list of FeatureMonitorJob resources for a specified FeatureMonitor resource, send a GET request by using the featureMonitorJobs.list method.

Before using any of the request data, make the following replacements:

  • LOCATION_ID: Region where the Feature resource is located, such as us-central1.
  • PROJECT_ID: Your project ID.
  • FEATUREGROUP_NAME: The name of the feature group containing the FeatureMonitor resource.
  • FEATURE_MONITOR_NAME: The name of the FeatureMonitor resource for which you want to list the feature monitoring jobs.

HTTP method and URL: 

GET https://LOCATION_ID-aiplatform.googleapis.com/v1beta1/projects/PROJECT_ID/locations/LOCATION_ID/featureGroups/FEATUREGROUP_NAME/featureMonitors/FEATURE_MONITOR_NAME/featureMonitorJobs

To send your request, choose one of these options:

curl

Execute the following command: 

curl -X GET \
     -H "Authorization: Bearer $(gcloud auth print-access-token)" \
     "https://LOCATION_ID-aiplatform.googleapis.com/v1beta1/projects/PROJECT_ID/locations/LOCATION_ID/featureGroups/FEATUREGROUP_NAME/featureMonitors/FEATURE_MONITOR_NAME/featureMonitorJobs"

PowerShell

Execute the following command: 

$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred" }

Invoke-WebRequest `
    -Method GET `
    -Headers $headers `
    -Uri "https://LOCATION_ID-aiplatform.googleapis.com/v1beta1/projects/PROJECT_ID/locations/LOCATION_ID/featureGroups/FEATUREGROUP_NAME/featureMonitors/FEATURE_MONITOR_NAME/featureMonitorJobs" | Select-Object -Expand Content

You should receive a JSON response similar to the following:

{
  "featureMonitorJobs": [
    {
      "name": "projects/PROJECT_NUMBER/locations/LOCATION_ID/featureGroups/FEATUREGROUP_NAME/featureMonitors/FEATURE_MONITOR_NAME/featureMonitorJobs/FEATURE_MONITOR_JOB_ID_1",
      "createTime": "2024-12-18T19:18:18.077161Z",
      "finalStatus": {},
      "featureSelectionConfig": {
        "featureConfigs": [
          {
            "featureId": "feature_name_1",
            "driftThreshold": 0.2
          },
          {
            "featureId": "feature_name_2",
            "driftThreshold": 0.2
          }
        ]
      }
    },
    {
      "name": "projects/PROJECT_NUMBER/locations/LOCATION_ID/featureGroups/FEATUREGROUP_NAME/featureMonitors/FEATURE_MONITOR_NAME/featureMonitorJobs/FEATURE_MONITOR_JOB_ID_2",
      "createTime": "2024-12-19T19:18:30.859921Z",
      "finalStatus": {},
      "featureSelectionConfig": {
        "featureConfigs": [
          {
            "featureId": "feature_name_1",
            "driftThreshold": 0.2
          },
          {
            "featureId": "feature_name_2",
            "driftThreshold": 0.2
          }
        ]
      }
    }
  ]
}

Python

Before trying this sample, follow the Python setup instructions in the Vertex AI quickstart using client libraries. For more information, see the Vertex AI Python API reference documentation.

To authenticate to Vertex AI, set up Application Default Credentials. For more information, see Set up authentication for a local development environment. 

from google.cloud import aiplatform
from vertexai.resources.preview import feature_store

aiplatofrm.init(project="PROJECT_ID", location="LOCATION_ID")

feature_group = FeatureGroup.get("FEATUREGROUP_NAME")
feature_monitor = feature_group.get_feature_monitor(FEATURE_MONITOR_NAME)
feature_monitor_jobs = feature_monitor.list_feature_monitor_jobs()

Replace the following:

  • LOCATION_ID: Region where the Feature resource is located, such as us-central1.
  • PROJECT_ID: Your project ID.
  • FEATUREGROUP_NAME: The name of the feature group containing the FeatureMonitor resource.
  • FEATURE_MONITOR_NAME: The name of the FeatureMonitor resource for which you want to list the feature monitoring jobs.

View feature statistics from a monitoring job

The following samples show how to view the feature statistics for all the features in a feature monitoring job. For each feature, the statistics and anomalies are displayed in the FeatureNameStatistics format.

REST

To view the feature statistics from a monitoring job by retrieving a FeatureMonitorJob resource, send a GET request by using the featureMonitorJobs.get method.

Before using any of the request data, make the following replacements:

  • LOCATION_ID: Region where where the feature monitoring job was run, such as us-central1.
  • PROJECT_ID: Your project ID.
  • FEATUREGROUP_NAME: The name of the feature group containing the FeatureMonitor resource.
  • FEATURE_MONITOR_NAME: The name of the FeatureMonitor resource for which the feature monitoring job was run.
  • FEATURE_MONITOR_JOB_ID: The ID of the FeatureMonitorJob resource that you want to retrieve.

HTTP method and URL: 

GET https://LOCATION_ID-aiplatform.googleapis.com/v1beta1/projects/PROJECT_ID/locations/LOCATION_ID/featureGroups/FEATUREGROUP_NAME/featureMonitors/FEATURE_MONITOR_NAME/featureMonitorJobs/FEATURE_MONITOR_JOB_ID

To send your request, choose one of these options:

curl

Execute the following command: 

curl -X GET \
     -H "Authorization: Bearer $(gcloud auth print-access-token)" \
     "https://LOCATION_ID-aiplatform.googleapis.com/v1beta1/projects/PROJECT_ID/locations/LOCATION_ID/featureGroups/FEATUREGROUP_NAME/featureMonitors/FEATURE_MONITOR_NAME/featureMonitorJobs/FEATURE_MONITOR_JOB_ID"

PowerShell

Execute the following command: 

$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred" }

Invoke-WebRequest `
    -Method GET `
    -Headers $headers `
    -Uri "https://LOCATION_ID-aiplatform.googleapis.com/v1beta1/projects/PROJECT_ID/locations/LOCATION_ID/featureGroups/FEATUREGROUP_NAME/featureMonitors/FEATURE_MONITOR_NAME/featureMonitorJobs/FEATURE_MONITOR_JOB_ID" | Select-Object -Expand Content

You should receive a JSON response similar to the following:

{
  "name": "projects/PROJECT_NUMBER/locations/LOCATION_ID/featureGroups/FEATUREGROUP_NAME/featureMonitors/FEATURE_MONITOR_NAME/featureMonitorJobs/FEATURE_MONITOR_JOB_ID",
  "createTime": "2024-12-19T19:18:18.077161Z",
  "finalStatus": {},
  "jobSummary": {
    "featureStatsAndAnomalies": [
      {
        "featureId": "feature_id_1",
        "featureStats": {
          "name": "feature_name_1",
          "type": "STRING",
          "stringStats": {
            "commonStats": {
              "numNonMissing": "6",
              "minNumValues": "1",
              "maxNumValues": "1",
              "avgNumValues": 1,
              "numValuesHistogram": {
                "buckets": [
                  {
                    "lowValue": 1,
                    "highValue": 1,
                    "sampleCount": 0.6
                  },
                  {
                    "lowValue": 1,
                    "highValue": 1,
                    "sampleCount": 0.6
                  }
                ],
                "type": "QUANTILES"
              },
              "totNumValues": "6"
            },
            "unique": "2",
            "topValues": [
              {
                "value": "59",
                "frequency": 2
              },
              {
                "value": "19",
                "frequency": 1
              }
            ],
            "avgLength": 2,
            "rankHistogram": {
              "buckets": [
                {
                  "label": "59",
                  "sampleCount": 2
                },
                {
                  "lowRank": "1",
                  "highRank": "1",
                  "label": "19",
                  "sampleCount": 1
                }
              ]
            }
          }
        },
        "statsTime": "2024-12-19T19:18:18.077161Z",
        "featureMonitorJobId": "FEATURE_MONITOR_JOB_ID",
        "featureMonitorId": "FEATURE_MONITOR_NAME"
      },
      {
        "featureId": "feature_id_2",
        "featureStats": {
          "name": "feature_name_1",
          "type": "STRING",
          "stringStats": {
            "commonStats": {
              "numNonMissing": "6",
              "minNumValues": "1",
              "maxNumValues": "1",
              "avgNumValues": 1,
              "numValuesHistogram": {
                "buckets": [
                  {
                    "lowValue": 1,
                    "highValue": 1,
                    "sampleCount": 0.6
                  },
                  {
                    "lowValue": 1,
                    "highValue": 1,
                    "sampleCount": 0.6
                  }
                ],
                "type": "QUANTILES"
              },
              "totNumValues": "6"
            },
            "unique": "2",
            "topValues": [
              {
                "value": "59",
                "frequency": 2
              },
              {
                "value": "19",
                "frequency": 1
              }
            ],
            "avgLength": 2,
            "rankHistogram": {
              "buckets": [
                {
                  "label": "59",
                  "sampleCount": 2
                },
                {
                  "lowRank": "1",
                  "highRank": "1",
                  "label": "19",
                  "sampleCount": 1
                }
              ]
            }
          }
        },
        "statsTime": "2024-12-19T19:18:18.077161Z",
        "featureMonitorJobId": "FEATURE_MONITOR_JOB_ID",
        "featureMonitorId": "FEATURE_MONITOR_NAME"
      }
    ]
  },
  "driftBaseFeatureMonitorJobId": "2250003330000300000",
  "driftBaseSnapshotTime": "2024-12-12T16:00:01.211686Z",
  "featureSelectionConfig": {
    "featureConfigs": [
      {
        "featureId": "feature_id_1",
        "driftThreshold": 0.2
      },
      {
        "featureId": "feature_id_2",
        "driftThreshold": 0.2
      }
    ]
  },
  "triggerType": "FEATURE_MONITOR_JOB_TRIGGER_ON_DEMAND"
}

Python

Before trying this sample, follow the Python setup instructions in the Vertex AI quickstart using client libraries. For more information, see the Vertex AI Python API reference documentation.

To authenticate to Vertex AI, set up Application Default Credentials. For more information, see Set up authentication for a local development environment. 

from google.cloud import aiplatform
from vertexai.resources.preview import feature_store

aiplatofrm.init(project="PROJECT_ID", location="LOCATION_ID")
feature_group = FeatureGroup.get("FEATUREGROUP_NAME"})

feature_monitor = feature_group.get_feature_monitor("FEATURE_MONITOR_NAME")
feature_monitor_job = feature_monitor.get_feature_monitor_job("FEATURE_MONITOR_JOB_ID)")

# Retrieve feature stats and anomalies
feature_stats_and_anomalies = feature_monitor_job.feature_stats_and_anomalies
print(feature_stats_and_anomalies)

Replace the following:

  • LOCATION_ID: Region where where the feature monitoring job was run, such as us-central1.
  • PROJECT_ID: Your project ID.
  • FEATUREGROUP_NAME: The name of the feature group containing the FeatureMonitor resource.
  • FEATURE_MONITOR_NAME: The name of the FeatureMonitor resource for which the feature monitoring job was run.
  • FEATURE_MONITOR_JOB_ID: The ID of the FeatureMonitorJob resource that you want to retrieve.

View feature statistics for a feature

You can retrieve the feature statistics for a specific feature from the most recent feature monitoring jobs executed, by retrieving the feature details and specifying the number of monitoring jobs that you want to retrieve the statistics from. The statistics and anomalies are displayed in the FeatureNameStatistics format.

The following samples show how to view the feature statistics for a specific feature from a specified number of recent feature monitoring jobs.

REST

To view the feature statistics for a specific feature in a Feature resource, send a GET request using the features.get method and specifying the number of monitoring jobs to retrieve the statistics from.

Before using any of the request data, make the following replacements:

  • LOCATION_ID: Region where where the feature monitoring job was run, such as us-central1.
  • PROJECT_ID: Your project ID.
  • FEATUREGROUP_NAME: The name of the feature group containing the feature.
  • FEATURE_NAME: The name of the Feature resource for which you want to retrieve the feature statistics.
  • LATEST_STATS_COUNT: The number of the latest monitoring jobs to retrieve the feature statistics from.

HTTP method and URL: 

GET https://LOCATION_ID-aiplatform.googleapis.com/v1beta1/projects/PROJECT_ID/locations/LOCATION_ID/featureGroups/FEATUREGROUP_NAME/features/FEATURE_NAME?feature_stats_and_anomaly_spec.latest_stats_count=LATEST_STATS_COUNT

To send your request, choose one of these options:

curl

Execute the following command: 

curl -X GET \
     -H "Authorization: Bearer $(gcloud auth print-access-token)" \
     "https://LOCATION_ID-aiplatform.googleapis.com/v1beta1/projects/PROJECT_ID/locations/LOCATION_ID/featureGroups/FEATUREGROUP_NAME/features/FEATURE_NAME?feature_stats_and_anomaly_spec.latest_stats_count=LATEST_STATS_COUNT"

PowerShell

Execute the following command: 

$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred" }

Invoke-WebRequest `
    -Method GET `
    -Headers $headers `
    -Uri "https://LOCATION_ID-aiplatform.googleapis.com/v1beta1/projects/PROJECT_ID/locations/LOCATION_ID/featureGroups/FEATUREGROUP_NAME/features/FEATURE_NAME?feature_stats_and_anomaly_spec.latest_stats_count=LATEST_STATS_COUNT" | Select-Object -Expand Content

You should receive a JSON response similar to the following:

{
  "name": "projects/PROJECT_NUMBER/locations/LOCATION_ID/featureGroups/FEATUREGROUP_NAME/features/FEATURE_NAME",
  "createTime": "2024-12-19T21:17:23.373559Z",
  "updateTime": "2024-12-19T21:17:23.373559Z",
  "etag": "sample_etag",
  "featureStatsAndAnomaly": [
    {
      "featureStats": {
        "name": "FEATURE_NAME",
        "type": "STRING",
        "stringStats": {
          "commonStats": {
            "numNonMissing": "4",
            "minNumValues": "1",
            "maxNumValues": "1",
            "avgNumValues": 1,
            "numValuesHistogram": {
              "buckets": [
                {
                  "lowValue": 1,
                  "highValue": 1,
                  "sampleCount": 0.4
                },
                {
                  "lowValue": 1,
                  "highValue": 1,
                  "sampleCount": 0.4
                },
                {
                  "lowValue": 1,
                  "highValue": 1,
                  "sampleCount": 0.4
                },
                {
                  "lowValue": 1,
                  "highValue": 1,
                  "sampleCount": 0.4
                }
              ],
              "type": "QUANTILES"
            },
            "totNumValues": "4"
          },
          "unique": "4",
          "topValues": [
            {
              "value": "feature_value_1",
              "frequency": 1
            },
            {
              "value": "feature_value_2",
              "frequency": 1
            },
            {
              "value": "feature_value_3",
              "frequency": 1
            },
            {
              "value": "feature_value_4",
              "frequency": 1
            }
          ],
          "avgLength": 4,
          "rankHistogram": {
            "buckets": [
              {
                "label": "label_1",
                "sampleCount": 1
              },
              {
                "lowRank": "1",
                "highRank": "1",
                "label": "label_2",
                "sampleCount": 1
              },
              {
                "lowRank": "2",
                "highRank": "2",
                "label": "label_3",
                "sampleCount": 1
              },
              {
                "lowRank": "3",
                "highRank": "3",
                "label": "label_4",
                "sampleCount": 1
              }
            ]
          }
        }
      },
      "driftDetectionThreshold": 0.1,
      "statsTime": "2024-12-19T22:00:02.734796Z",
      "featureMonitorJobId": "feature_monitor_job_id_1",
      "featureMonitorId": "feature_monitor_name_1"
    }
  ],
  "versionColumnName": "version_column_name"
}

Python

Before trying this sample, follow the Python setup instructions in the Vertex AI quickstart using client libraries. For more information, see the Vertex AI Python API reference documentation.

To authenticate to Vertex AI, set up Application Default Credentials. For more information, see Set up authentication for a local development environment. 

from google.cloud import aiplatform
from vertexai.resources.preview import feature_store

aiplatofrm.init(project="PROJECT_ID", location="LOCATION_ID")

feature_group = FeatureGroup.get("FEATUREGROUP_NAME"})
feature_stats_and_anomalies = feature_group.get_feature("FEATURE_NAME", latest_stats_count=LATEST_STATS_COUNT)
print(feature_stats_and_anomalies)

Replace the following:

  • LOCATION_ID: Region where where the feature monitoring job was run, such as us-central1.
  • PROJECT_ID: Your project ID.
  • FEATUREGROUP_NAME: The name of the feature group containing the FeatureMonitor resource.
  • FEATURE_NAME: The name of the feature for which you want to retrieve the feature statistics.
  • LATEST_STATS_COUNT: The number of latest monitoring jobs to retrieve the feature statistics from.

Example use case: Use feature monitoring to detect feature drift

You can use feature monitoring to detect an anomaly in feature data called feature drift. A drift is a significant and unforeseen change to feature data in BigQuery over time. Vertex AI Feature Store helps you identify feature drift by comparing the snapshot at the time when the monitoring job is run, with the data snapshot during the previous monitoring job execution.

For any feature included in the feature monitor, if the difference between the two snapshots exceeds the threshold specified in the drift_threshold parameter, Vertex AI Feature Store identifies a feature drift and returns the following information in the FeatureMonitorJob resource:

  • The driftDetected parameter is set to true.

  • The distribution deviation between the two snapshots. For numerical features, Vertex AI Feature Store calculates this value using Jensen-Shannon divergence. For categorical features, Vertex AI Feature Store calculates this value using L-infinity distance.

  • The threshold that was exceeded by the drift score.

The following samples show how to retrieve a FeatureMonitorJob resource and verify whether a drift was detected.

REST

To retrieve a FeatureMonitorJob resource , send a GET request by using the featureMonitorJobs.get method.

Before using any of the request data, make the following replacements:

  • LOCATION_ID: Region where where the feature monitoring job was run, such as us-central1.
  • PROJECT_ID: Your project ID.
  • FEATUREGROUP_NAME: The name of the feature group containing the FeatureMonitor resource.
  • FEATURE_MONITOR_NAME: The name of the FeatureMonitor resource for which the feature monitoring job was run.
  • FEATURE_MONITOR_JOB_ID: The ID of the FeatureMonitorJob resource that you want to retrieve.

HTTP method and URL: 

GET https://LOCATION_ID-aiplatform.googleapis.com/v1beta1/projects/PROJECT_ID/locations/LOCATION_ID/featureGroups/FEATUREGROUP_NAME/featureMonitors/FEATURE_MONITOR_NAME/featureMonitorJobs/FEATURE_MONITOR_JOB_ID

To send your request, choose one of these options:

curl

Execute the following command: 

curl -X GET \
     -H "Authorization: Bearer $(gcloud auth print-access-token)" \
     "https://LOCATION_ID-aiplatform.googleapis.com/v1beta1/projects/PROJECT_ID/locations/LOCATION_ID/featureGroups/FEATUREGROUP_NAME/featureMonitors/FEATURE_MONITOR_NAME/featureMonitorJobs/FEATURE_MONITOR_JOB_ID"

PowerShell

Execute the following command: 

$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred" }

Invoke-WebRequest `
    -Method GET `
    -Headers $headers `
    -Uri "https://LOCATION_ID-aiplatform.googleapis.com/v1beta1/projects/PROJECT_ID/locations/LOCATION_ID/featureGroups/FEATUREGROUP_NAME/featureMonitors/FEATURE_MONITOR_NAME/featureMonitorJobs/FEATURE_MONITOR_JOB_ID" | Select-Object -Expand Content

You should receive a JSON response similar to the following:

{
  "name": "projects/PROJECT_NUMBER/locations/LOCATION_ID/featureGroups/FEATUREGROUP_NAME/featureMonitors/FEATURE_MONITOR_NAME/featureMonitorJobs/FEATURE_MONITOR_JOB_ID",
  "createTime": "2024-12-14T19:45:30.026522Z",
  "finalStatus": {},
  "jobSummary": {
    "featureStatsAndAnomalies": [
      {
        "featureId": "feature_id_1",
        "featureStats": {
          "name": "feature_name_1",
          "type": "STRING",
          "stringStats": {
            "commonStats": {
              "numNonMissing": "3",
              "minNumValues": "1",
              "maxNumValues": "1",
              "avgNumValues": 1,
              "numValuesHistogram": {
                "buckets": [
                  {
                    "lowValue": 1,
                    "highValue": 1,
                    "sampleCount": 0.9
                  },
                  {
                    "lowValue": 1,
                    "highValue": 1,
                    "sampleCount": 0.9
                  },
                  {
                    "lowValue": 1,
                    "highValue": 1,
                    "sampleCount": 0.9
                  }
                ],
                "type": "QUANTILES"
              },
              "totNumValues": "3"
            },
            "unique": "3",
            "topValues": [
              {
                "value": "sample_value_1",
                "frequency": 1
              },
              {
                "value": "sample_value_2",
                "frequency": 1
              },
              {
                "value": "sample_value_3",
                "frequency": 1
              }
            ],
            "avgLength": 3,
            "rankHistogram": {
              "buckets": [
                {
                  "label": "sample_label_1",
                  "sampleCount": 1
                },
                {
                  "lowRank": "1",
                  "highRank": "1",
                  "label": "sample_label_2",
                  "sampleCount": 1
                },
                {
                  "lowRank": "2",
                  "highRank": "3",
                  "label": "sample_label_3",
                  "sampleCount": 1
                }
              ]
            }
          }
        },
        "distributionDeviation": 0.1388880008888000,
        "driftDetectionThreshold": 0.1,
        "driftDetected": true,
        "statsTime": "2024-12-15T19:45:37.026522Z",
        "featureMonitorJobId": "FEATURE_MONITOR_JOB_ID",
        "featureMonitorId": "FEATURE_MONITOR_NAME"
      }
    ]
  },
  "driftBaseFeatureMonitorJobId": "2250003330000300000",
  "driftBaseSnapshotTime": "2024-12-12T18:18:18.077161Z",
  "description": "sample_feature_monitor_job_description",
  "featureSelectionConfig": {
    "featureConfigs": [
      {
        "featureId": "feature_name",
        "driftThreshold": 0.1
      }
    ]
  },
  "triggerType": "FEATURE_MONITOR_JOB_TRIGGER_ON_DEMAND"
}

Python

Before trying this sample, follow the Python setup instructions in the Vertex AI quickstart using client libraries. For more information, see the Vertex AI Python API reference documentation.

To authenticate to Vertex AI, set up Application Default Credentials. For more information, see Set up authentication for a local development environment. 

from google.cloud import aiplatform
from vertexai.resources.preview import feature_store

aiplatofrm.init(project="PROJECT_ID", location="LOCATION_ID")
feature_group = FeatureGroup.get("FEATUREGROUP_NAME"})
feature_monitor = feature_group.get_feature_monitor("FEATURE_MONITOR_NAME")
feature_monitor_job = feature_monitor.get_feature_monitor_job("FEATURE_MONITOR_JOB_ID)")

# Retrieve feature stats and anomalies
feature_stats_and_anomalies = feature_monitor_job.feature_stats_and_anomalies
print(feature_stats_and_anomalies)

# Check whether drifts are detected
for feature_stats_and_anomalies in feature_monitor_job.feature_stats_and_anomalies:
    print("feature: ", feature_stats_and_anomalies.feature_id)
    print("drift score: ", feature_stats_and_anomalies.distribution_deviation)
    print("drift detected: ", feature_stats_and_anomalies.drift_detected)

Replace the following:

  • LOCATION_ID: Region where where the feature monitoring job was run, such as us-central1.
  • PROJECT_ID: Your project ID.
  • FEATUREGROUP_NAME: The name of the feature group containing the FeatureMonitor resource.
  • FEATURE_MONITOR_NAME: The name of the FeatureMonitor resource for which the feature monitoring job was run.
  • FEATURE_MONITOR_JOB_ID: The ID of the FeatureMonitorJob resource that you want to retrieve.