How can you explain this appar 12. Under what conditions would we be able to use expected profit as an indicator of the relative performance of classification models? (Or, alternatively, think of the conditions under which we would not be able to use expected profit or expected profit would be suspect.)