This version of the Bayesian classifier is the most commonly studied and used, because of its easy implementation, but especially because of the very good results obtained in most cases.

However, this method involves a loss of accuracy because of the assumption of conditional independence.

This operator also makes it possible to merge agreeing information. The product of the probabilities assigned by the various criteria to a same element corresponds to a very strict fusion where all of the criteria must be satisfied, at least a little, so that the considered class can be selected. We further develop the concepts of fusion with conjunctive behaviour (strict) or disjunctive behaviour (cautious).