libsurv

Introduction

A library of efficient survival analysis models, including DeepCox, HitBoost, BecCox 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.
BecCox: Adding convex function approximated concordance index in EfnBoost to adjust risk ranking.

Enhancement

comprehensive document
python package distribution

Installation

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

Usage

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

Hyper-parameters tuning can refer to libsurv/bysopt.

Computation acceleration

As the objective function gradients are computed using numpy package, the model fitting round would be much slow and need to be optimized.

User can use pytorch to finish array or matrix computation instead of numpy. Furthermore, you can benefit from the power of GPUs. What is not acceptable is that you need to implement it yourself.

Citation

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

HitBoost(Accepted by IEEE-Access): P. Liu, B. Fu and S. X. Yang, "HitBoost: Survival Analysis via a Multi-Output Gradient Boosting Decision Tree Method," in IEEE Access, vol. 7, pp. 56785-56795, 2019, doi: 10.1109/ACCESS.2019.2913428.
EfnBoost(Accepted by IEEE-TBME): P. Liu, B. Fu, S. X. Yang, L. Deng, X. Zhong and H. Zheng, "Optimizing Survival Analysis of XGBoost for Ties to Predict Disease Progression of Breast Cancer," in IEEE Transactions on Biomedical Engineering, doi: 10.1109/TBME.2020.2993278.
DeepCox(Under Review): Deep Survival Learning for Predicting the Overall Survival in Breast Cancer using Clinical and Follow-up Data

liupei101 / libsurv