The Research Of Chinese Spam Filtering Technology Based On Na(?)ve Bayes Classification Algorithm

With the rapid popularization of computer and the development of the Internet, Email service has become the most popular communication tool to most computer users. But the occurrence of Junk Email not only seriously disturbs and impacts the normal use of mail system, but also impedes the further development of Internet.At the beginning, this dissertation analyzes the defects of current email system and the reason of Junk email flooding; narrates current spam situation and the related measures and technology used to fight against spam; compares various spam filter algorithms, especially analyzes the one based on Naive Bayesian.This dissertation studies the general framework of spam filter; analyzes the differences of spam classifier between on English mails and Chinses mails; introduces the N-gram algorithm to resolve Chinese character segment processing. There still are some defects and limitations about the mail classifier based on Naive Bayes algorithm, such as the limitation of independent assumption etc, this dissertation proposes some ideas to improve the Naive Bayes algorithm. Based on a deep investigation on the spam corpus, this dissertation proposes a 2-layer classification structure based on mail length. In this dissertation, a Chinese mail classifier based on the improved Naive Bayes algorithm is designed and some experiments is made to verify it.In the end, this dissertation also proposes a 2-layer filtering structure which combinates the rule based technique and the statistic based technique, and realizes the prototype of this structure. The related experiments show that this structure can embody the user's preference.