I am using autoML for training purposes, however my dataset is very imbalanced. It is a multiclass task with 5 classes and I want to minimize this effect when training. I have verified the documentation and it seems autoML support the creation of a new column with weights.
My doubts are: What is the range I am supposed to use? At this moment I am using 30, 40, 1.5 and 4 (I imagine weight should be used as multiplier). Can I use them? About new dataset, I know this column will not exist anymore, so how can the model understand that?
Length of each class:
A: 12730 B: 181733 C: 623848 D: 1056696 E: 18210
