STLDetector<T>

Models & Types · AiDotNet.AnomalyDetection.TimeSeries

Detects anomalies using STL (Seasonal and Trend decomposition using Loess).

Inheritance: ModelBase<T, Matrix<T>, Vector<T>> → AnomalyDetectorBase<T> → STLDetector<T>

Implements: IGeneratedParameterRegistrar<T>

For Beginners

STL decomposes a time series into three components: trend, seasonal, and residual. The residual component contains the irregular variations. Large residuals indicate anomalies that don’t fit the expected trend and seasonal patterns.

How It Works

The algorithm works by:

  1. Apply STL decomposition to extract trend, seasonal, and residual
  2. Standardize the residual component
  3. Large standardized residuals indicate anomalies

When to use:

  • Time series with clear trend and/or seasonal patterns
  • When anomalies disrupt expected patterns
  • Sales, weather, sensor data with periodicity

Industry Standard Defaults:

  • Season length: 7 (weekly pattern)
  • Trend smoothness: 15
  • Contamination: 0.1 (10%)

Reference: Cleveland, R.B., et al. (1990). “STL: A Seasonal-Trend Decomposition Procedure Based on Loess.” Journal of Official Statistics.

Constructors

Constructor Summary
STLDetector(int seasonLength, int trendSmoothness, double contamination, int randomSeed) Creates a new STL anomaly detector.

Properties

Property Summary
int SeasonLength { get; } Gets the season length.
int TrendSmoothness { get; } Gets the trend smoothness parameter.

Methods

Method Summary
Vector<T> ScoreAnomalies(Matrix<T> X)
void Fit(Matrix<T> X)