MovingAverageDetector<T>

Models & Types · AiDotNet.AnomalyDetection.TimeSeries

Detects anomalies using moving average deviation in time series.

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

For Beginners

This detector computes a moving average and identifies points that deviate significantly from their local average. It’s one of the simplest and most intuitive time series anomaly detection methods.

How It Works

The algorithm works by:

  1. Compute moving average and moving standard deviation
  2. For each point, compute deviation from local mean in std units
  3. High deviations indicate anomalies

When to use:

  • Simple time series without complex patterns
  • As a baseline method
  • Real-time anomaly detection (streaming data)

Industry Standard Defaults:

  • Window size: 20
  • Threshold: 3 standard deviations
  • Contamination: 0.1 (10%)

Constructors

Constructor Summary
MovingAverageDetector(int windowSize, double stdThreshold, double contamination, int randomSeed) Creates a new Moving Average anomaly detector.

Properties

Property Summary
double StdThreshold { get; } Gets the standard deviation threshold.
int WindowSize { get; } Gets the window size.

Methods

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