I have a very fundamental question about dependent variable, which is not like ‘price’ but more like a label but having very high cardinality. It’s like some ID of a row or ‘product_id’ in true sense. This product_id can range from 1-100K.
- Can we have ‘Y’ as label/ID with such large value for prediction? We can surly use log function but the back conversion to absolute value may not produce exact ID.
- Do we need to have that large FC layer for this to get resolved?
Any hint would be beneficially. Thanks!