search

sktime / skpro

A unified framework for tabular probabilistic regression, time-to-event prediction, and probability distributions in python
https://skpro.readthedocs.io/en/latest
BSD 3-Clause "New" or "Revised" License
244 stars 45 forks source link

[ENH] `ngboost` survival prediction model #290

Closed fkiraly closed 5 months ago

fkiraly commented 6 months ago

The PR https://github.com/sktime/skpro/pull/215 has added ngboost as an skpro regressor.

The package can also be invoked for survival prediction, i.e., for using censoring information C as in the skpro survival models. We should also add a version of the ngboost algorithm for survival prediction, hence.

To avoid user confusion, it might be better to add this as a separate estimator in the survival module, although the two estimators may end up sharing a lot of similarities and hence, possibly, an inheritance tree.

ShreeshaM07 commented 6 months ago

Should I start working on this issue?

fkiraly commented 6 months ago

If you find it interesting, of course!

Given that you've interfaced ngboost first time round, it should be less work for you than anyone.

fkiraly commented 6 months ago

Perhaps the most useful examples of recently interfaced survival models:

https://github.com/sktime/skpro/pull/247 https://github.com/sktime/skpro/pull/237

The main additional challenges are: