Survival analysis with any regressor

[thierrymoudiki]

Hi,

I implemented this (work in progress): a way to use any regressor (including in boosting) having fit and predict.

https://github.com/thierrymoudiki/scikit-survival/blob/master/examples/2024-11-14-scikit-survival.ipynb

What's your opinion? Is this something that could be on your agenda, or should I follow my own route?

T.

[sebp]

Hello and thanks for reaching out.

Can you explain a little bit what your thought process was to allow for other regressors? Is this based on a particular paper?

[thierrymoudiki]

Sure. Here's my thought process:

1. First, I thought it was possible (because I use to do similar things for supervised learning and time series forecasting)
2. I searched for an 'entry point' for a long time, that I found in your Gradient Boosting Analysis, and replaced the base learner by any model having fit and predict
3. For the Custom class, I relied on the Gradient Boosting implementation: only one boosting iteration is used, and a learning rate = 1.0

Not based on a paper at all (yet, but could be in the future)

[thierrymoudiki]

Follow-up: survivalist: Probabilistic model-agnostic survival analysis using scikit-learn, glmnet, xgboost, lightgbm, pytorch, keras, nnetsauce and mlsauce

https://thierrymoudiki.github.io/blog/2024/12/15/python/agnostic-survival-analysis