Survivors is a Python library for survival analysis designed to handle real-world, messy, and heterogeneous data with high accuracy and interpretability.

Unlike many existing libraries that rely on strong theoretical assumptions (e.g., proportional hazards, exponential distributions), Survivors is:

• Flexible: works with incomplete, noisy, and heterogeneous tabular data.
• Interpretable: powered by survival trees and ensembles (forests) that offer transparent decision paths.
• Highly sensitive to data structure: tree splits are adapted to handle informative censoring and arbitrary time-to-event distributions.
• Accurate: avoids restrictive model assumptions, resulting in higher predictive performance in complex settings.

Key Features

Survivors was developed during a PhD project at Lomonosov Moscow State University to overcome the limitations of traditional survival analysis. It is designed to handle data with informative censoring and arbitrary time-to-event distributions. Here's what makes it stand out:

• Data Included – Comes with 20+ real-world datasets from healthcare, CRM, and reliability.
• True Continuous-Time Modeling – No discretization needed; time is continuous where it matters: at prediction.
• Custom Metrics That Matter – Over 10 evaluation metrics crafted to handle censoring, imbalance, and rare events.
• Speed Meets Scale – Optimized from scratch with JIT compilation, vectorized math, and histogram-based splits for efficient performance.
• Experiment-Ready – Built-in module for reproducible experiments: hold-out, cross-validation, time-aware splits, and hyperparameter search.
• Meta-Modeling Power – Create stratified models and ensemble aggregators with ease.
• Broad Applicability – Suitable for a wide range of domains including healthcare (death or discharge prediction), industrial maintenance (equipment failure), and business analytics (customer churn and retention).

Installation

Install the latest release via pip (PyPI):

pip install survivors

Or install the development version:

git clone https://github.com/iuliivasilev/dev-survivors.git
cd dev-survivors
pip install -e .

Dependencies

• Python >= 3.8
• NumPy, Pandas, Numba, Scikit-learn, Matplotlib

Optional for comprehensive experiments:

• lifelines (>= 0.27.8)
• scikit-survival (>= 0.17.2)

Getting Started

The user guides provide detailed information on the key concepts for survivors:

import survivors.datasets as ds
import survivors.constants as cnt
from survivors.tree import CRAID

X, y, features, categ, sch_nan = ds.load_gbsg_dataset()

cr = CRAID(criterion='wilcoxon', depth=2, min_samples_leaf=0.1,
        signif=0.05, categ=categ, leaf_model="base")
cr.fit(X, y)
tree_view = cr.visualize(mode="surv")
tree_view

Supported Models

• CRAID – interpretable decision tree model for survival analysis. [1]
• ParallelBootstrapCRAID – ensemble of independent trees with improved stability and accuracy [2]
• BoostingCRAID (optional) – boosting ensemble with adaptive sampling [3]
• Modified existing models (Kaplan-Meier, CoxPH, AFT ...)

Help and Support

We welcome contributions! Open issues, propose features, and submit pull requests via GitHub.

For questions, discussions, or collaboration ideas, feel free to open an issue or contact the maintainer directly (email).

Citation

If you use survivors in a scientific publication, we would appreciate citations:

[1]	Vasilev I., Petrovskiy M., Mashechkin I. Survival Analysis Algorithms based on Decision Trees with Weighted Log-rank Criteria. - 2022.

[2]	Vasilev, Iulii, Mikhail Petrovskiy, and Igor Mashechkin. "Sensitivity of Survival Analysis Metrics." Mathematics 11.20 (2023): 4246.

[3]	Vasilev, Iulii, Mikhail Petrovskiy, and Igor Mashechkin. "Adaptive Sampling for Weighted Log-Rank Survival Trees Boosting." International Conference on Pattern Recognition Applications and Methods. Cham: Springer International Publishing, 2021.