search

loft-br / xgboost-survival-embeddings

Improving XGBoost survival analysis with embeddings and debiased estimators
https://loft-br.github.io/xgboost-survival-embeddings/
Apache License 2.0
321 stars 53 forks source link

Cure Fraction Problem? #35

Closed NudnikShpilkis closed 3 years ago

NudnikShpilkis commented 3 years ago

What is the "cure fraction" problem described in the documentation for the XGBSEStackedWeibull. To quote: We also have better extrapolation capabilities, as opposed to the cure fraction problem in `XGBSEKaplanNeighbors` and `XGBSEKaplanTree`.

GabrielGimenez commented 3 years ago

XGBSEKaplanNeighbors and XGBSEKaplanTreecan't iterate over times that has no sample, so if you dataset has a high cure fraction you might end over the median even at the last time.

XGBSEStackedWeibull is a parametric method, so it can predict even at times you dataset has no observations, so we say it can naturally extrapolates your results to the time intervals that you want