Toward Conditional Distribution Calibration in Survival Prediction

  • 2025-02-03 16:47:09
  • Shi-ang Qi, Yakun Yu, Russell Greiner
  • 0

Abstract

Survival prediction often involves estimating the time-to-event distributionfrom censored datasets. Previous approaches have focused on enhancingdiscrimination and marginal calibration. In this paper, we highlight thesignificance of conditional calibration for real-world applications --especially its role in individual decision-making. We propose a method based onconformal prediction that uses the model's predicted individual survivalprobability at that instance's observed time. This method effectively improvesthe model's marginal and conditional calibration, without compromisingdiscrimination. We provide asymptotic theoretical guarantees for both marginaland conditional calibration and test it extensively across 15 diversereal-world datasets, demonstrating the method's practical effectiveness andversatility in various settings.

 

Quick Read (beta)

loading the full paper ...