FeatureBaggingDetector<T>
Models & Types · AiDotNet.AnomalyDetection.Ensemble
Detects anomalies using Feature Bagging ensemble method.
Inheritance: ModelBase<T, Matrix<T>, Vector<T>> → AnomalyDetectorBase<T> → FeatureBaggingDetector<T>
Implements: IGeneratedParameterRegistrar<T>
For Beginners
Feature Bagging creates multiple anomaly detectors, each trained on a random subset of features. The final score is the combination of all detector scores. This helps handle high-dimensional data where different feature subsets may reveal different anomalies.
How It Works
The algorithm works by:
- Create n_estimators base detectors
- Each detector uses a random subset of features
- Train each detector on its feature subset
- Combine scores using averaging or maximum
When to use:
- High-dimensional data
- When different feature combinations may reveal different anomalies
- When you want more robust detection than single-detector methods
Industry Standard Defaults:
- N estimators: 10
- Max features: 0.5 (50% of features per detector)
- Contamination: 0.1 (10%)
Reference: Lazarevic, A., Kumar, V. (2005). “Feature Bagging for Outlier Detection.” KDD.
Constructors
| Constructor | Summary |
|---|---|
FeatureBaggingDetector(int nEstimators, double maxFeatures, CombinationMethod<T> combination, double contamination, int randomSeed) |
Creates a new Feature Bagging anomaly detector. |
Properties
| Property | Summary |
|---|---|
double MaxFeatures { get; } |
Gets the maximum features fraction. |
int NEstimators { get; } |
Gets the number of estimators. |
Methods
| Method | Summary |
|---|---|
Vector<T> ScoreAnomalies(Matrix<T> X) |
|
void Fit(Matrix<T> X) |