I have a dataset containing binary labels Y with values "correct" or "wrong".
I want to study the reason of wrong state, based on several conditions made from multiple features.
For example I want to know if "wrong" was caused by a "common wrong answer" or by "difficulty to answer", i.e. I want to get Y with 3 labels "correct ", "common wrong" and "difficult".
Is this possible to make by using machine learning model such as logistic regression ? If not, is there any viable approach to this problem?