FairCutForest<T>
Models & Types · AiDotNet.AnomalyDetection.TreeBased
Implements Fair-Cut Forest (FCF) for anomaly detection with balanced tree construction.
Inheritance: ModelBase<T, Matrix<T>, Vector<T>> → AnomalyDetectorBase<T> → FairCutForest<T>
For Beginners
Fair-Cut Forest improves on Isolation Forest by using balanced tree construction. Instead of random splits, it selects splits that more evenly divide the data, leading to more consistent isolation of anomalies.
How It Works
The algorithm works by:
- Building trees with “fair” splits that balance the data more evenly
- Using importance-weighted feature selection based on data spread
- Computing anomaly scores based on average path length to isolation
When to use:
- High-dimensional data where standard Isolation Forest struggles
- When you need more consistent anomaly rankings
- Datasets with features of varying importance
Industry Standard Defaults:
- Number of trees: 100
- Max samples: 256
- Contamination: 0.1 (10%)
Reference: Inspired by improvements to Isolation Forest for fair/balanced splits.
Constructors
| Constructor | Summary |
|---|---|
FairCutForest(int numTrees, int maxSamples, double contamination, int randomSeed) |
Creates a new Fair-Cut Forest anomaly detector. |
Properties
| Property | Summary |
|---|---|
int MaxSamples { get; } |
Gets the maximum samples used per tree. |
int NumTrees { get; } |
Gets the number of trees in the forest. |
Methods
| Method | Summary |
|---|---|
Vector<T> ScoreAnomalies(Matrix<T> X) |
|
void Fit(Matrix<T> X) |