Abstract: Using a classifier based on a specific machinclcarning technique to automatically filter out spam email has drawn many researchers' attention. In a spam filtering process,how to selecting the features of emails and how to design a good filtering algorithm arc two key issues. A new method of features selecting was proposed, which include the head and the other main features of emails. Furthermore, the features' importance degree was measured according to information viewpoint of rough set. With it,a new weighted naW a I3ayes spam filtering was put forward. It can solve the conditional dependence of naW c Bayes efficiently. Simulation results on two email data sets in English and Chinese respectively illustrate the efficiency of this method.

Key words: Spam filtering, Feature selecting, Rough set, Weighted naive Bayes