HurstExponentSelector<T>
Models & Types · AiDotNet.Preprocessing.FeatureSelection.Complexity
Hurst Exponent based Feature Selection.
Inheritance: TransformerBase<T, Matrix<T>, Matrix<T>> → HurstExponentSelector<T>
Implements: IFixedCountFeatureSelector<T>
For Beginners
The Hurst exponent H indicates trend behavior: H > 0.5 means trending (persistent) - ups followed by ups, downs by downs; H < 0.5 means mean-reverting (anti-persistent) - ups followed by downs; H = 0.5 means random walk (no memory). Financial time series with H ≠ 0.5 may contain exploitable patterns.
How It Works
Selects features based on their Hurst exponent, which measures long-range dependence and self-similarity in time series.
Constructors
| Constructor | Summary |
|---|---|
HurstExponentSelector(int nFeaturesToSelect, bool preferTrending, int[] columnIndices) |
Properties
| Property | Summary |
|---|---|
Vector<T> HurstValues { get; } |
|
bool PreferTrending { get; } |
|
bool SupportsInverseTransform { get; } |
|
int NFeaturesToSelect { get; } |
|
int[] SelectedIndices { get; } |
Methods
| Method | Summary |
|---|---|
Matrix<T> FitTransform(Matrix<T> data) |
|
bool[] GetSupportMask() |
|
string[] GetFeatureNamesOut(string[] inputFeatureNames) |