GitHub Repository

Scalable, scikit-learn-compatible competing-risks survival analysis in pure Python — CR random survival forest, Fine-Gray, cause-specific Cox, Aalen-Johansen CIF, Gray's test, and exact TreeSHAP. 10–22× faster than randomForestSRC on real EHR and 16.6–544× vs scikit-survival (n=5k→50k); scales to n=10⁶ in ~1 min.

7 starsPython

Crforest – Competing-risks RSF in Python, 6× faster than R's rfSRC

Name: Crforest – Competing-risks RSF in Python, 6× faster than R's rfSRC
Availability: InStock
Author: sunnyadn

by sunnyadn·Apr 29, 2026·1 point·0 comments

Visit Project View on HN

AI Analysis

●●●BangerNiche GemSolve My Problem

Native Python competing-risks RSF that's 6x faster than R's randomForestSrc.

Strengths

•Bit-identical reproducibility mode enables direct R migration without recalculating any baseline results.
•Scales to one million rows in minutes where R's randomForestSRC currently fails.
•Scikit-learn compatible API reduces friction for existing standard Python machine learning workflows.

Weaknesses

•Pre-alpha status means API instability until the promised stable version 1.0 release.
•Highly niche domain limits adoption outside specific epidemiology and reliability engineering fields.