Skip to content

CRISP-NAM: Competing Risks for Interpretable Survival Analysis using Neural Additive Models

This repository contains research code for the paper: CRISP-NAM: Competing Risks Interpretable Survival Prediction with Neural Additive Models. It includes the Python code for the following:

  • Models: CRISP-NAM and DeepHIT
  • Data loading utilities for 4 datasets: Framingham, PBC, Support2, Synthetic
  • Training scripts: Standard training, Hyperparameter optimization via Optuna, Nested cross validation
  • Metrics: Loss and risk functions for survival analysis.
  • Plotting: Feature importance and Shape functions for interpretability.

PyPI package

The core files of research: models, metrics and plotting utilities.

Installation

You can install the package via the following pip command: 

pip install crisp_nam

Citation

@inproceedings{ramachandram2025crispnam,
title={CRISP-NAM: Competing Risks Interpretable Survival Prediction with Neural Additive Models},
author={Ramachandram, Dhanesh and Raval, Ananya},
booktitle={EXPLIMED 2025 - Second Workshop on Explainable AI for the Medical Domain},
year={2025}
}