Bayesian Calibration

Calibration API

Fine-tune risk thresholds based on historical outcomes. Bayesian updating for smarter scoring.

How Calibration Works

OmegaEngine uses Bayesian calibration to continuously improve risk scoring accuracy. As you provide feedback on decisions, the system updates its priors and adjusts thresholds.

P(risky | features) = P(features | risky) × P(risky) / P(features)

Updated with each graded decision

Calibration Metrics

Brier Score

Measures prediction accuracy

0.12

ECE

Expected Calibration Error

0.04

API Reference

// Get current calibration state
const calibration = await omega.calibration.get({
  policyId: "default"
});

// Trigger recalibration
await omega.calibration.recalibrate({
  policyId: "default",
  windowDays: 30,
  minSamples: 100
});

// Set custom thresholds
await omega.calibration.setThresholds({
  policyId: "default",
  escalateAbove: 0.75,
  blockAbove: 0.90
});