Toward Conditional Distribution Calibration in Survival Prediction

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Qi, Shi-ang, Yu, Yakun, Greiner, Russell
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916660054589440
author Qi, Shi-ang
Yu, Yakun
Greiner, Russell
author_facet Qi, Shi-ang
Yu, Yakun
Greiner, Russell
contents Survival prediction often involves estimating the time-to-event distribution from censored datasets. Previous approaches have focused on enhancing discrimination and marginal calibration. In this paper, we highlight the significance of conditional calibration for real-world applications -- especially its role in individual decision-making. We propose a method based on conformal prediction that uses the model's predicted individual survival probability at that instance's observed time. This method effectively improves the model's marginal and conditional calibration, without compromising discrimination. We provide asymptotic theoretical guarantees for both marginal and conditional calibration and test it extensively across 15 diverse real-world datasets, demonstrating the method's practical effectiveness and versatility in various settings.
format Preprint
id arxiv_https___arxiv_org_abs_2410_20579
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Toward Conditional Distribution Calibration in Survival Prediction
Qi, Shi-ang
Yu, Yakun
Greiner, Russell
Machine Learning
Artificial Intelligence
Survival prediction often involves estimating the time-to-event distribution from censored datasets. Previous approaches have focused on enhancing discrimination and marginal calibration. In this paper, we highlight the significance of conditional calibration for real-world applications -- especially its role in individual decision-making. We propose a method based on conformal prediction that uses the model's predicted individual survival probability at that instance's observed time. This method effectively improves the model's marginal and conditional calibration, without compromising discrimination. We provide asymptotic theoretical guarantees for both marginal and conditional calibration and test it extensively across 15 diverse real-world datasets, demonstrating the method's practical effectiveness and versatility in various settings.
title Toward Conditional Distribution Calibration in Survival Prediction
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2410.20579