search

suiji / Arborist

Scalable decision tree training and inference.
Other
82 stars 14 forks source link

Feature request: Include OOB error as an attribute of the 'validation' object #19

Closed talgalili closed 7 years ago

talgalili commented 7 years ago

Is it possible to get the out of bag error of the model? I can't seem to find this in the object or any function that gets it.

(if not - then I think this is an important feature to include)

suiji commented 7 years ago

Right now it would have to be computed from the residuals. We can easily include it as an attribute of the 'validation' object, in future.

talgalili commented 7 years ago

That would be very helpful, thanks.

suiji commented 7 years ago

Just now getting to this - there were squeakier wheels to oil.

For classification, we will report OOB error as the mean misprediction rate.

For regression, we already report MSE and R-squared. Would you like to see a mean error rate, as well?

talgalili commented 7 years ago

I think the MSE is probably fine, although it never hearts to add another error measure.

suiji commented 7 years ago

Not a problem. Should be available soon.

suiji commented 7 years ago

Validation now includes oob error, as discussed. Regression now reports mean absolute error.

Please feel free to re-open this item if other issues arise.