Scikit-Survival Analysis

关于

This skill equips Claude with specialized knowledge for performing survival analysis in Python, leveraging the scikit-survival library built on top of scikit-learn. It enables the creation of sophisticated time-to-event models—ranging from traditional Cox proportional hazards and penalized regression to advanced machine learning approaches like Random Survival Forests and Gradient Boosting. The skill provides expert guidance on handling the unique challenges of censored data, analyzing competing risks, and implementing specialized evaluation metrics such as Uno's C-index and Brier scores to ensure robust predictive performance in clinical, industrial, or business contexts.

主要功能

• Non-parametric estimation using Kaplan-Meier and Nelson-Aalen estimators
• Specialized survival modeling including Cox PH, Random Survival Forests, and Gradient Boosting
• Competing risks analysis for handling multiple mutually exclusive event outcomes
• Advanced evaluation metrics for censored data like Concordance Index and Time-Dependent AUC
• 1 GitHub stars
• Comprehensive data preprocessing for right, left, and interval-censored datasets

使用场景

• Modeling customer churn and time-to-attrition for subscription-based business models
• Analyzing patient survival rates and identifying risk factors in clinical and medical research
• Predicting time-to-failure and performing predictive maintenance for industrial equipment