SurvivalTree with min_samples_leaf based on observed events

[tomer1812]

Hi,

I'm attempting to prevent node splitting in a SurvivalTree when there are fewer than 6 observed events in either the right or left nodes after a split.
Using min_samples_leaf didn't achieve the desired outcome as it accounts for all samples, not just the observed events.
I also tried min_weight_fraction_leaf with the observed event indicator column as sample_weight, but this approach didn't work and resulted in a smaller number of observed events in each leaf.
How can I properly configure the scikit-survival SurvivalTree to ensure a minimum number of observed events in each leaf?

Thank you!

[sebp]

Unfortunately, there is currently no way to achieve this. Also note that using sample_weight is currently broken (see #443).