Intelligent Detection Approaches for Spam

This paper proposes intelligent detection approaches based on incremental support vector machine and artificial immune system for the spam of e-mail stream. In the approaches, a window is used to hold several classifiers each of which classifies the e-mail independently and the label of the e-mail is given by a strategy of majority voting. Exceeding margin update technique is also used for the dynamical update of each classifier in the window. A sliding window is employed for purge of out-of-date knowledge so far. Techniques above endow our algorithm with dynamical and adaptive properties as well as the ability to trace the changing of the content of e-mails and user's interests in a continuous way. We conduct many experiments on two public benchmark corpus called PU1 and Ling. Experimental results demonstrate that the proposed intelligent detection approaches for spam give a promising performance.