The target variable is discrete values
- means classes
- the values take on only a small number of discrete values
- classification function is defined by
Data Classifications
Data Classification Notion
It's not that regression is bad, but that classification is better as the final evaluation loss function, according to Google Deepmind