Really interesting! If I may suggest a slight change, for the confidence classification it would be useful to have probability percentages or a probability yardstick