QuantileDispersionSelector<T>
Models & Types · AiDotNet.Preprocessing.FeatureSelection.Quantile
Quantile Dispersion (QCD) based Feature Selection.
Inheritance: TransformerBase<T, Matrix<T>, Matrix<T>> → QuantileDispersionSelector<T>
Implements: IFixedCountFeatureSelector<T>
For Beginners
QCD measures spread relative to the central location, similar to coefficient of variation but using quartiles instead of mean/std. It’s robust to outliers and works well when data contains extreme values. Values range from 0 (no spread) to 1 (maximum relative spread).
How It Works
Selects features based on the Quartile Coefficient of Dispersion, a robust measure of relative variability: QCD = (Q3 - Q1) / (Q3 + Q1)
Constructors
| Constructor | Summary |
|---|---|
QuantileDispersionSelector(int nFeaturesToSelect, int[] columnIndices) |
Properties
| Property | Summary |
|---|---|
Vector<T> QCDValues { 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) |