Abstract: Bayesian filtering algorithm is one of most widely used methods of anti-spam filtering technology. Taking into account the fact that the wrong classification of the mail causes different losses to recipients, so introducing a message that if judging as a spam mail is λ times that of judging as a normal mail, it can conclude that this is a spam mail. Meanwhile, in order to improve the quality of classification, the paper uses genetic algorithm to distinguish between tokens in the body and tokens in the subject. Finally, using the sample to validate the improved algorithm, the result shows that using new algorithm can improve the quality of the message classification.

Key words: Bayesian, anti-spam mail, genetic algorithm