libsurv

Introduction

A library of efficient survival analysis models, including DeepCox, HitBoost, CEBoost and EfnBoost methods.

DeepCox: Deep cox proportional hazard model implemented by tensorflow. It's exactly the same as TFDeepSurv.
HitBoost: Survival analysis via a multi-output gradient boosting decision tree method.
EfnBoost: Optimized cox proportional hazard model via an objective function of Efron approximation.
CEBoost: Adding convex function approximated concordance index in EfnBoost to adjust risk ranking.

# in the directory where `setup.py` is located
ls
# install via pip or pip3 (only support for python>=3.5)
pip3 install .

Usage of DeepCox, EfnBoost, CEBoost and HitBoost are provided in Jupyter Notebooks.

Hyper-parameters tuning can refer to libsurv/bysopt.

If you would like to cite our package, some reference papers are listed below:

HitBoost(Accepted by IEEE-Access): HitBoost: Survival Analysis via a Multi-output Gradient Boosting Decision Tree method.
EfnBoost(Accepted by IEEE-TBME): Optimizing Survival Analysis of XGBoost for Ties to Predict Prognostic Status of Breast Cancer
DeepCox(Under Review): Deep Survival Learning for Predicting the Overall Survival in Breast Cancer using Clinical and Follow-up Data