JamesSteinEncoder<T>

Models & Types ยท AiDotNet.Preprocessing.Encoders

Encodes categorical features using James-Stein shrinkage estimation.

Inheritance: TransformerBase<T, Matrix<T>, Matrix<T>> โ†’ JamesSteinEncoder<T>

For Beginners

This encoder balances between:

  • Trusting category-specific averages (when we have lots of data)
  • Falling back to the overall average (when category data is sparse)
  • The balance is determined automatically using statistical theory

How It Works

JamesSteinEncoder uses Bayesian shrinkage to blend category-specific target means with the global mean. Categories with more samples get weights closer to their own mean, while rare categories shrink toward the global mean.

The shrinkage formula: encoded = (1 - B) * category_mean + B * global_mean where B is the shrinkage factor based on sample size and variance.

Constructors

Constructor Summary
JamesSteinEncoder(JamesSteinHandleUnknown handleUnknown, int[] columnIndices) Creates a new instance of JamesSteinEncoder.

Properties

Property Summary
JamesSteinHandleUnknown 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.

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 James-Stein shrinkage estimates.