Bootstrap632PlusSplitter<T>

Models & Types ยท AiDotNet.Preprocessing.DataPreparation.Splitting.Bootstrap

.632+ Bootstrap splitter that improves upon .632 Bootstrap for high-variance scenarios.

Inheritance: DataSplitterBase<T> โ†’ Bootstrap632PlusSplitter<T>

For Beginners

The .632 bootstrap can still be biased when the model overfits. The .632+ method adaptively adjusts the weighting based on how much the model overfits, providing more reliable estimates.

How It Works

How It Works: The .632+ error is calculated as: error_632+ = (1-w)error_train + werror_test where w varies between 0.632 and 1 based on overfitting degree.

Mathematical Details:

  • R = (error_test - error_train) / (gamma - error_train)
  • gamma = error under the null model (random predictions)
  • w = 0.632 / (1 - 0.368 * R)

Constructors

Constructor Summary
Bootstrap632PlusSplitter(int nIterations, int randomSeed) Creates a new .632+ Bootstrap splitter.

Properties

Property Summary
int NumSplits { get; }
string Description { get; }

Methods

Method Summary
DataSplitResult<T> Split(Matrix<T> X, Vector<T> y)
IEnumerable<DataSplitResult<T>> GetSplits(Matrix<T> X, Vector<T> y)