CalibrationOptions

Options & Configuration · AiDotNet.Evaluation.Options

Configuration options for probability calibration analysis.

For Beginners

When your model says “80% chance this email is spam”, calibration checks if it’s really spam 80% of the time. Poorly calibrated models might say 80% but actually be right 95% or only 50% of the time. Good calibration is essential when you use predicted probabilities for decisions.

How It Works

Calibration measures whether predicted probabilities match actual frequencies. A well-calibrated model that predicts 80% probability should be correct 80% of the time.

Constructors

Constructor Summary
CalibrationOptions()

Properties

Property Summary
BinningStrategy? BinningStrategy { get; set; } Binning strategy.
ProbabilityCalibrationMethod? CalibrationMethod { get; set; } Calibration method to apply.
bool? ComputeAdaptiveECE { get; set; } Whether to compute Adaptive ECE.
bool? ComputeBrierDecomposition { get; set; } Whether to compute Brier score decomposition.
bool? ComputeCalibrationSlope { get; set; } Whether to compute calibration slope/intercept.
bool? ComputeConfidenceIntervals { get; set; } Whether to compute confidence intervals for calibration metrics.
bool? ComputeECE { get; set; } Whether to compute Expected Calibration Error (ECE).
bool? ComputeMCE { get; set; } Whether to compute Maximum Calibration Error (MCE).
bool? ComputePerClassCalibration { get; set; } Whether to compute per-class calibration.
bool? GenerateConfidenceHistogram { get; set; } Whether to generate confidence histogram data.
bool? GenerateReliabilityDiagram { get; set; } Whether to generate reliability diagram data.
bool? RunHosmerLemeshowTest { get; set; } Whether to run Hosmer-Lemeshow test.
bool? RunSpiegelhalterTest { get; set; } Whether to run Spiegelhalter’s z-test.
bool? UseCrossValidation { get; set; } Whether to use cross-validation for calibration metrics.
double? ConfidenceLevel { get; set; } Confidence level.
double? RegularizationStrength { get; set; } Regularization strength for calibrator fitting.
double? Temperature { get; set; } Temperature for temperature scaling calibration.
int? BootstrapSamples { get; set; } Number of bootstrap samples for CIs.
int? CVFolds { get; set; } Number of CV folds if using cross-validation.
int? MaxIterations { get; set; } Maximum iterations for calibrator optimization.
int? NumberOfBins { get; set; } Number of bins for reliability diagram.