CalibratedClassifierOptions<T>

Options & Configuration · AiDotNet.Models.Options

Configuration options for calibrated classifiers.

Inheritance: ModelOptions → ClassifierOptions<T> → CalibratedClassifierOptions<T>

For Beginners

Calibration improves the probability estimates from classifiers.

Choose your method based on your needs:

  • Platt Scaling: Good for SVMs and linear models. Assumes S-shaped transformation.
  • Isotonic Regression: Non-parametric, very flexible. Best with lots of data.
  • Beta Calibration: More flexible than Platt, handles asymmetric distortions.
  • Temperature Scaling: Simple and effective for neural networks.

Key settings:

  • Use cross-validation (CrossValidationFolds > 1) for better calibration with less data
  • More CV folds = better calibration but slower training
  • Set Seed for reproducibility

Constructors

Constructor Summary
CalibratedClassifierOptions()

Properties

Property Summary
ProbabilityCalibrationMethod CalibrationMethod { get; set; } Gets or sets the calibration method.
double CalibrationSetFraction { get; set; } Gets or sets the fraction of data to use for calibration when not using cross-validation.
int CrossValidationFolds { get; set; } Gets or sets the number of cross-validation folds.