search

zachmayer / caretEnsemble

caret models all the way down :turtle:
Other
226 stars 75 forks source link

Is the option metric = 'logLoss' available in train()? #203

Closed randel closed 8 years ago

randel commented 8 years ago

With train() in the caret package, there is an option metric = "logLoss". But for caretEnsemble, using this option will always trigger the warning below. It seems that this option has been blocked? Any feedback would be really appreciated.

The metric "logLoss" was not in the result set. ROC will be used instead.

zachmayer commented 8 years ago

You need to pass a custom summaryFunction to trainControl that calculates log loss.

By default, caret::train only computes accuracy.

randel commented 8 years ago

Thanks for your help! For anyone interested, here is a workable function for log loss.

zachmayer commented 8 years ago

No problem, and thanks for the link!