GBMSurv: Second-Order Gradient Boosting for Survival Analysis

GBMSurv is a collection of gradient boosting-based models for survival analysis. It leverages gradient boosting techniques to predict survival times and perform risk analysis for censored survival data. The model is particularly useful for predicting time-to-event data, which is common in medical research, reliability analysis, and various engineering applications. This repository contains implementations of all model variants.

Installation

To ensure compatibility and avoid conflicts, it is recommended to set up an isolated Python environment. You can use conda for this purpose.

To install GBMSurv in development mode after cloning the repository, follow these steps:

git clone https://github.com/NTAILab/GBMSurv.git
cd GBMSurv
pip install -e .

Package Contents

The repository is organized into several main directories:

gbmsurv/

The main package directory. It includes shared utilities and three sub-packages for different gradient boosting survival model variants:

• gbmsurv/log_norm_gbmsurv/ – parametric model based on the Log-Normal distribution.
• gbmsurv/weib_gbmsurv/ – parametric model based on the Weibull distribution.
• gbmsurv/non_param_gbmsurv/ – non-parametric model that estimates event probabilities within discrete time intervals.

Each model subpackage contains the following modules:

• model.py – defines the core gradient boosting survival model for that variant.
• loss.py – implements the loss function and its gradient/Hessian logic, adapted for survival analysis.
• __init__.py – imports the main model class for easy access.

The root package also includes:

• utils.py – shared utility functions for data preprocessing, metric computations, and experimental support.
• __init__.py – aggregates all model variants for simplified import

examples/

Contains example notebooks demonstrating how to train and evaluate each survival model variant on survival data.

Usage

Example usage is provided in the examples directory, including a demonstration of the model's application to survival datasets.

To use the model for survival analysis, follow these steps:

1. Preprocess the dataset, ensuring it contains censored survival times (e.g., time-to-event data) in the format (delta, time) where:

   • delta: Censoring indicator (1 if the event occurred, 0 if the data is censored).
   • time: The observed survival time. The target variable y should be in the form of a structured NumPy array with the fields delta and time.

2. Define the required model using NonParamGBMSurv, LogNormGBMSurv or WeibGBMSurv.

3. Train the model and evaluate performance metrics, such as the C-index, for model evaluation.

Here’s an example of using GBMSurv models for survival analysis:

from gbmsurv import (
    WeibullGBMSurvModel,
    LogNormGBMSurvModel,
    NonParamGBMSurvModel,
)

from sksurv.datasets import load_veterans_lung_cancer
from sklearn.model_selection import train_test_split

X, y = load_veterans_lung_cancer()
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

weib = WeibGBMSurvivalModel(n_estimators=100, max_depth=5, learning_rate=0.1)
lognorm = LogNormGBMSurvivalModel(n_estimators=100, max_depth=5, learning_rate=0.1)
gbmsurv = GBMSurvivalModel(n_estimators=100, max_depth=5, learning_rate=0.1)

weib.fit(X_train, y_train)
lognorm.fit(X_train, y_train)
gbmsurv.fit(X_train, y_train)

preds_weib = weib.predict(X_test)
preds_lognorm = lognorm.predict(X_test)
preds_gbmsurv = gbmsurv.predict(X_test)

c_index_weib = weib.score(X_test, y_test)
c_index_lognorm = lognorm.score(X_test, y_test)
c_index_gbmsurv = gbmsurv.score(X_test, y_test)

print(f'C-index (Weibull): {c_index_weib}')
print(f'C-index (Log-Normal): {c_index_lognorm}')
print(f'C-index (Non-Parametric): {c_index_gbmsurv}')

This will train the NonParamGBMSurv, LogNormGBMSurv and WeibGBMSurv models on Veterans dataset and provide predictions on test data.

Citation

If you use this project in your research, please cite it as follows:

...will be later.