search

harrysouthworth / gbm

Gradient boosted models
Other
106 stars 27 forks source link

Bernoulli requires the response to be in {0,1}... #29

Closed Mullefa closed 9 years ago

Mullefa commented 9 years ago

... and similarly for other distributions. Whilst this doesn't affect the functionality of the package, it is a bit inconvenient.

Could the requirement be lowered to factors with two levels, and implicit conversion done within gbm.fit() (or whatever function from within the package is being used)?

Likewise, in the correspondingpredict() method, could a factor be returned instead of 0's and 1's?

ghost commented 9 years ago

+++

harrysouthworth commented 9 years ago

This issue was moved to gbm-developers/gbm#6