Jupytercon 2023

JupyterCon 2023 - Predictive Survival Analysis and Competing Risk Modeling

jupytercon_sk

See the poster of the talk (opens in a new tab)

This tutorial is split into 2 notebooks:

  • Part 1: we introduce the main metrics and estimators used in survival analysis, along with our new Gradient Boosted CIF model for competing events.

  • Part 2: we detail the concept of sessionization: by transforming raw logs of activity, we can create labels for survival analysis in an implicit or heartbeat fashion. We showcase this feature engineering with both DuckDB and Polars.

Repo of the demo

https://github.com/soda-inria/survival-analysis-benchmark (opens in a new tab)