MEstimateEncoder<T>

Models & Types · AiDotNet.Preprocessing.Encoders

Encodes categorical features using M-estimate regularization.

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

For Beginners

M-estimate is like adding ‘m’ fake samples:

  • Each fake sample has the global mean as its target
  • Categories with few samples get pulled toward the global mean
  • Categories with many samples stay close to their actual mean
  • Higher m = more smoothing toward global mean

How It Works

MEstimateEncoder applies M-estimate smoothing to target encoding, which adds a regularization parameter ‘m’ that controls shrinkage toward the global mean.

The formula: encoded = (n * category_mean + m * global_mean) / (n + m) where n is the count of samples in the category and m is the smoothing parameter.

Constructors

Constructor Summary
MEstimateEncoder(double m, MEstimateHandleUnknown handleUnknown, int[] columnIndices) Creates a new instance of MEstimateEncoder.

Properties

Property Summary
MEstimateHandleUnknown HandleUnknown { get; } Gets how unknown categories are handled.
T GlobalMean { get; } Gets the global target mean.
bool SupportsInverseTransform { get; } Gets whether this transformer supports inverse transformation.
double M { get; } Gets the smoothing parameter m.

Methods

Method Summary
Matrix<T> FitTransform(Matrix<T> data, Vector<T> target) Fits and transforms the data.
string[] GetFeatureNamesOut(string[] inputFeatureNames) Gets the output feature names after transformation.
void Fit(Matrix<T> data, Vector<T> target) Fits the encoder by computing M-estimate encodings.