Learning possibilistic networks from data: a survey