I have a data frame which I want to use for multi class classification problem. There are total 6 classes (say a, b, c, d, e, f). I want to improve the precision for three classes (say a, b, c) i.e. model should only predict one of the three classes if it's absolutely certain that the sample won't lie in either of three other classes (i.e. in either of d, e, f).
I know I can introduce imbalance or use another binary classifier. But I want to do this task using a skewed loss function. What properties should I consider in this loss function? Also I don't see how can I put custom loss function in catboost algorithm?
