In Risk, I am seeing 100s of rules deployed for example one where they look at sales in the last 30days > 5000 and model score > 0.5 then ‘suspend account’ ,etc I know that they use decision tree to create these rules. But nowhere do they mention what the y variable used for the decision tree is. How can you go about defining the Y label? Is there a method to this?

Like can you use ‘account suspended = Y/N in 60 days’ That itself as the Y label and then predict? Or do I have to look at other metrics like default rate > x% for instance and predict who they are and then suspend them? I understand it is specific to the problem but what is the idea behind choosing a Y label for a decision tree problem in the real world?