摘要
Spam, the electronic equivalent of junk mail, affects over 600 million users worldwide. The massive increase of spam is posing a very serious threat to email which has become an important means of communication. Not only does it annoy users, but it also consumes much of the bandwidth of the Internet. Even as anti-spam solutions change to limit the amount of spam sent to users, the senders adapt to make sure their messages are seen, while spam reorganization has different properties comparing with normal text reorganization. Presented are three different classifiers are combined with detailed analysis on various training data set of the given spam database. These classifiers are combined into a mixture of experts (MOE) system which yields overall better performance than any of the individual contributors. The instructions for further improvements on classifiers as well as its requirement on spam databases are also given.
Spam, the electronic equivalent of junk mail, affects over 600 million users worldwide. The massive increase of spam is posing a very serious threat to email which has become an important means of communication. Not only does it annoy users, but it also consumes much of the bandwidth of the Internet. Even as anti-spam solutions change to limit the amount of spam sent to users, the senders adapt to make sure their messages are seen, while spam reorganization has different properties comparing with normal text reorganization. Presented are three different classifiers are combined with detailed analysis on various training data set of the given spam database. These classifiers are combined into a mixture of experts (MOE) system which yields overall better performance than any of the individual contributors. The instructions for further improvements on classifiers as well as its requirement on spam databases are also given.
