Support vector machines for spam categorization：（支持向量机分类为垃圾邮件）.pdfVIP

1048 IEEE TRANSACTIONS ON NEURAL NETWORKS, VOL. 10, NO. 5, SEPTEMBER 1999 Support Vector Machines for Spam Categorization Harris Drucker, Senior Member, IEEE, Donghui Wu, Student Member, IEEE, and Vladimir N. Vapnik Abstract— We study the use of support vector machines the classifier) of being nonspam is at the top of the list. In (SVM’s) in classifying e-mail as spam or nonspam by comparing either of these modes, the filter does not reject any messages, it to three other classification algorithms: Ripper, Rocchio, and only indicates whether the message has a high priority of being boosting decision trees. These four algorithms were tested on two different data sets: one data set where the number of features spam. were constrained to the 1000 best features and another data set It is highly desirable that if the user decides that e-mail where the dimensionality was over 7000. SVM’s performed best messages be rank-ordered by degree of confidence that the when using binary features. For both data sets, boosting trees rank ordering be reliable. By reliable, we mean that the user and SVM’s had acceptable test performance in terms of accuracy can either start at the top of the list of e-mail messages and speed. However, SVM’s had significantly less training time. and be fairly confident that they represent nonspam messages Index Terms— Boosting algorithms, classification, e-mail, fea- or start at the bottom of the list and be confident that the ture representation, Ripper, Rocchio, support vector machines. messages are spam. It is only near the middle of the list