메시지 규칙을 이용한 나이브 베이지안 스팸메일 필터링 시스템

초록

This paper suggests hybrid spam-mail filtering system is based on message rules and NBC(Naive Bayesian Classifier) which is used widely in existing document classification. This system can classify spam-mails automatically and doesn't need users to make rules. After useless elements is removed from new mail through document preprocessing, that mail is filtered through naive bayesian filter if it isn't filtered through message rules based filter. In experiment, we evaluate this system compared with message rules based system and naive bayesian classifier system for proving performance of the proposed system. Also, the performance of this system can be improved as changing threshold.