Online Filtering Method and Optimization Based on SMO and E-mail Fingerprint
Bin Chen · Journal of Zhengzhou University · 2009
The finger features vectoring and the SVM filtering method are proposed on the spam,and an online SVM spam filter called FSVM is designed and implemented,which attains the corresponding performance as the classic method in the online spam filtering.In view of the computing speed of SVM filtering,a dynamic example set reduction method(DFSVM) is given out for SVM filtering method based on the original SMO algorithm,which can greatly reduce the computing cost and keep the corresponding performance.The experiment and comparison test running on the standard corpus is given out in the actual mail system.It is proved that the optimization can improve the accuracy of SVM classification.