ARIMADetector<T>
Models & Types · AiDotNet.AnomalyDetection.TimeSeries
Detects anomalies in time series using ARIMA model residuals.
Inheritance: ModelBase<T, Matrix<T>, Vector<T>> → AnomalyDetectorBase<T> → ARIMADetector<T>
Implements: IGeneratedParameterRegistrar<T>
For Beginners
ARIMA (AutoRegressive Integrated Moving Average) is a classic time series model. It predicts future values based on past values and errors. Points where the prediction error is large are flagged as anomalies.
How It Works
The algorithm works by:
- Fit an ARIMA(p,d,q) model to the time series
- Compute prediction residuals
- Points with large residuals (standardized) are anomalies
When to use:
- Stationary time series (or made stationary via differencing)
- Detecting point anomalies that don’t fit the temporal pattern
- Well-understood temporal dependencies
Industry Standard Defaults:
- p (AR order): 2
- d (differencing): 1
- q (MA order): 2
- Contamination: 0.1 (10%)
Reference: Box, G.E.P., Jenkins, G.M. (1970). “Time Series Analysis: Forecasting and Control.”
Constructors
| Constructor | Summary |
|---|---|
ARIMADetector(int p, int d, int q, double contamination, int randomSeed) |
Creates a new ARIMA anomaly detector. |
Properties
| Property | Summary |
|---|---|
int D { get; } |
Gets the differencing order (d). |
int P { get; } |
Gets the AR order (p). |
int Q { get; } |
Gets the MA order (q). |
Methods
| Method | Summary |
|---|---|
Vector<T> ScoreAnomalies(Matrix<T> X) |
|
void Fit(Matrix<T> X) |