With the continuous development of Internet, email has become a more popular forms of communication in people's life, and as well spam has become a major problem in E-mail users. Thus suppressing spam has become a rather urgent task. A minimum risk bayes algorithm based on the traditional bayesian is proposed on the basis of loss factor k, By adjusting the k value, the algorithm can improve the influence of spam false negative result, and reduce the loss of users at its best. At last, The results of the experiment indicates that the minimum risk bayes algorithm can make spam has a better filtering effect.