How do we find out what was the issue with machine learning model A , why model A dint work but a model B worked [closed]