Drift: Configs

Histogram Analyzer

Analyzing drift of numerical features. Note that we use group:continuous to select all continuous features.

config:
  algorithm: hellinger
  baseline:
    size: 14
    type: TrailingWindow
  metric: histogram
  minBatchSize: 1
  threshold: 0.7
  type: drift
id: numerical_drift-analyzer
schedule:
  cadence: daily
  type: fixed
targetMatrix:
  include:
  - group:continuous
  segments:
  - tags: []
  type: column

Frequent Items Analyzer

Analyzing frequent items to detect drifts with categorical values

config:
  algorithm: hellinger
  baseline:
    size: 14
    type: TrailingWindow
  metric: frequent_items
  minBatchSize: 1
  threshold: 0.7
  type: drift
id: frequent_items_drift-analyzer
schedule:
  cadence: daily
  type: fixed
targetMatrix:
  include:
  - group:discrete
  segments:
  - tags: []
  type: column

Histogram with custom threshold

Customize the threshold that will trigger the anomaly for a histogram.

config:
  algorithm: hellinger
  baseline:
    size: 14
    type: TrailingWindow
  metric: histogram
  minBatchSize: 1
  threshold: 0.5
  type: drift
id: drift-histogram-segments-analyzer
schedule:
  cadence: daily
  type: fixed
targetMatrix:
  include:
  - group:continuous
  type: column