A classification method including first classifying an event of any kind by first rules, and then second classifying events, not identified by the first classification, by a learning base reinforced with all the events identified by the first classification. The method is adaptive if the second classification rules are amended according to new examples that were able to be determined by the first rules.