The Python package cenreg is a repository for probabilistic forecasts such as quantile regression and distribution regression and for censored regression such as survival analysis and interval-censored data analysis.
Features:
- Tree-based models run fast and output accurate predictions.
- Neural network models are implemented both for structured data (e.g., tabular data) and non-structured data (e.g., image data).
- Both models can handle competing risks.
- Both models are based on the (conditional) independence assumption or the non-informative assuption, but they can also handle dependent censoring based on assumed copula.
- Strictly proper scoring rules are implemented to evaluate the discrimination performances of prediction models. The scoring rules can handle right-censored and interval-censored data.
- Binning-free calibration metrics are implemented to evaluate the calibration performances of prediction models. The calibration metrics can handle right-censored and interval-censored data.
You first need to install SurvSet via pip
pip install SurvSet
Additionally, denpending on the models you want to use, you also need to install
- LightGBM
- PyTorch
You can install cenreg via pip:
pip install cenreg
You can find our sample codes in the notebooks directory.
Read the documentation to get started.
- H. Yanagisawa and S. Akiyama, A Strictly Proper Scoring Rule and a Calibration Metric for Interval-Censored Data Analysis, ICML 2026 (Paper in OpenReview)
- H. Yanagisawa and S. Akiyama, Survival Analysis via Density Estimation, ICML 2025 (Paper in OpenReview)
- H. Yanagisawa, Proper Scoring Rules for Survival Analysis, ICML 2023