TargetEncoder<T>

Models & Types · AiDotNet.Preprocessing.Encoders

Encodes categorical features using target mean encoding.

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

For Beginners

Instead of one-hot encoding (many columns), target encoding creates a single column per feature containing the average target value for each category:

  • Category “A” with average target 0.8 becomes 0.8
  • Category “B” with average target 0.3 becomes 0.3

This is especially useful for high-cardinality features where one-hot would create too many columns.

How It Works

TargetEncoder replaces each category with the mean of the target variable for that category. This creates a continuous feature that captures the relationship between the category and target.

To prevent overfitting, especially with rare categories, smoothing is applied: encoding = (count * category_mean + smoothing * global_mean) / (count + smoothing)

Constructors

Constructor Summary
TargetEncoder(double smoothing, double minSamplesLeaf, TargetEncoderHandleUnknown handleUnknown, int[] columnIndices) Creates a new instance of TargetEncoder.

Properties

Property Summary
Dictionary<int, CategoryValueMap<T>> EncodingMaps { get; } Gets the encoding maps for each column.
TargetEncoderHandleUnknown HandleUnknown { get; } Gets how unknown categories are handled during transform.
bool SupportsInverseTransform { get; } Gets whether this transformer supports inverse transformation.
double Smoothing { get; } Gets the smoothing parameter used during encoding.

Methods

Method Summary
Matrix<T> FitTransform(Matrix<T> data, Vector<T> target) Fits the encoder and transforms the data in one step.
string[] GetFeatureNamesOut(string[] inputFeatureNames) Gets the output feature names after transformation.
void Fit(Matrix<T> data, Vector<T> target) Fits the encoder by learning the target means for each category.