What metrics should I use to evaluate survival models?

The skill guides you in using specialized metrics like the Concordance Index (C-index), Brier scores, and time-dependent AUC, which account for censoring better than standard regression metrics.

What is scikit-survival?

Scikit-survival is a Python library built on top of scikit-learn specifically designed for survival analysis, which involves predicting the time until an event of interest occurs.

Does this skill support non-linear survival relationships?

Yes, it provides implementation patterns for non-linear models including Random Survival Forests, Gradient Boosting, and Kernel-based Survival SVMs.

Which model should I use for high-dimensional survival data?

For datasets where the number of features is large relative to the number of samples, this skill recommends using CoxnetSurvivalAnalysis which applies elastic net regularization.

How does this skill help with censored data?

The skill provides standardized patterns for using the Surv utility to handle observations where the event time is only partially known, a common challenge in survival datasets.

Scikit-Survival for Python

Name: Scikit-Survival for Python
Author: x-cmd

byx-cmd

•

Ciencia de Datos y ML

Implements advanced survival analysis and time-to-event modeling using the scikit-survival library in Python.

This skill provides a comprehensive toolkit for survival analysis, enabling AI agents and developers to model time-to-event data effectively within the scikit-learn ecosystem. It offers specialized guidance on handling right-censored observations, fitting proportional hazards models, building ensemble survival forests, and evaluating model performance using domain-specific metrics like the concordance index and Brier score. By integrating best practices for data preparation and model selection, it streamlines the development of predictive models for clinical trials, customer churn analysis, and equipment reliability engineering.

Características Principales

01Survival-specific evaluation via Uno's C-index and Brier scores

02Advanced data preparation for right, left, and interval-censored data

03Ensemble methods including Random Survival Forests and Gradient Boosting

04Comprehensive support for Cox Proportional Hazards and penalized Coxnet models

05Competing risks analysis and non-parametric Kaplan-Meier estimation

068 GitHub stars

Casos de Uso

01Predicting patient survival outcomes and event risks in clinical research

02Modeling customer churn and time-to-attrition for subscription services

03Estimating time-to-failure for predictive maintenance in industrial engineering

Características Principales

01Survival-specific evaluation via Uno's C-index and Brier scores

02Advanced data preparation for right, left, and interval-censored data

03Ensemble methods including Random Survival Forests and Gradient Boosting

04Comprehensive support for Cox Proportional Hazards and penalized Coxnet models

05Competing risks analysis and non-parametric Kaplan-Meier estimation

068 GitHub stars

Casos de Uso

01Predicting patient survival outcomes and event risks in clinical research

02Modeling customer churn and time-to-attrition for subscription services

03Estimating time-to-failure for predictive maintenance in industrial engineering