Abstract: Email is favored by people and has become one of the most universal information communication methods, inspired by its convenience and low cost. However, E-mail has also been abused by malicious people. As a result, Internet has been polluted by spam. Spam not only wastes bandwidth resource, disrupting people's normal life and work, but also influences routine application of mail servers and poses a serious threat to Internet security. Decision tree algorithm was utilized to train extracted mail head feature information to construct a decision tree, which could be further used to filter spam. Variable precision rough set model was introduced to classify some specific instance or noise data into appropriate class, avoiding the situation when normal mails were regarded as junk mails, reducing losses bringing to users. As is shown in the result, the proposed algorithm is feasible in spam filtering and decreases the rate of normal mails being regarded as junk.

Key words: spam, filtering, feature information, variable precision rough set, decision tree