Predictions are currently made using a Zero-One Loss Function, which corresponds to a case where the cost associated with misclassifying each outcome is the same. There may be situations however (cancer diagnosis, for instance) where there is a higher misclassification cost with an outcome (in this case, the person having cancer).
I was going to include the possibility to define and use cost functions in my PR for Naive Bayes, but I think it is better to add the functionality with another PR.
Predictions are currently made using a Zero-One Loss Function, which corresponds to a case where the cost associated with misclassifying each outcome is the same. There may be situations however (cancer diagnosis, for instance) where there is a higher misclassification cost with an outcome (in this case, the person having cancer).
I was going to include the possibility to define and use cost functions in my PR for Naive Bayes, but I think it is better to add the functionality with another PR.