AnomalyFeaturesTransformer<T>

Models & Types ยท AiDotNet.Preprocessing.TimeSeries

Computes rolling anomaly detection features for time series data.

Inheritance: TimeSeriesTransformerBase<T> โ†’ AnomalyFeaturesTransformer<T>

For Beginners

Anomaly detection helps find unusual or unexpected values in your data.

This is useful for:

  • Fraud detection (unusual transactions)
  • Equipment failure prediction (abnormal sensor readings)
  • Quality control (out-of-specification products)
  • Financial monitoring (unusual market movements)

The transformer creates several types of anomaly features:

  • Z-score: How many standard deviations from normal
  • IQR outliers: Values outside the typical range
  • CUSUM: Detects gradual shifts in the data
  • Isolation score: Machine learning-based anomaly scoring

How It Works

This transformer calculates features that help identify unusual patterns, including Z-scores, IQR-based outlier detection, CUSUM control charts, and isolation scores.

Constructors

Constructor Summary
AnomalyFeaturesTransformer(TimeSeriesFeatureOptions options) Creates a new anomaly features transformer with the specified options.

Properties

Property Summary
bool SupportsInverseTransform { get; }