MCFS<T>

Models & Types · AiDotNet.Preprocessing.FeatureSelection.Filter.Spectral

Multi-Cluster Feature Selection using spectral graph analysis.

Inheritance: TransformerBase<T, Matrix<T>, Matrix<T>> → MCFS<T>

Implements: IFixedCountFeatureSelector<T>

For Beginners

MCFS finds features that help preserve the natural groupings (clusters) in your data. It first discovers hidden cluster structures using spectral methods, then picks features that best capture those clusters. Great when you don’t have labels but want features that preserve data structure.

How It Works

MCFS uses spectral analysis to find features that preserve the multi-cluster structure in data. It computes cluster indicators via spectral clustering and then selects features using sparse regression with L1 regularization.

Constructors

Constructor Summary
MCFS(int nFeaturesToSelect, int nClusters, int nNeighbors, double alpha, int? randomState, int[] columnIndices)

Properties

Property Summary
Vector<T> MCFSScores { 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)