E-mail Classification Algorithm Based On Dynamic Artificial Immune

E-mail is increasingly becoming one of the most convenient communication ways in modern world. Nevertheless, large quantity of spam mails are greedily swallowing the rare recourses of Internet and bringing about serious social problems. Yet the spam filtering technologies now being applied are defficientiate because of the limitation such as time-consuming training and insensitive to changes. Artificial immune system (AIS) is characterized by its excellent immune identification, learning , memory and its self/nonself discrimination ability, providing a solution for solving the spam E-mail problems.In this paper, an introduction of Artificial immune system is given including its origin, theoretical researches and related algorithms. And the deficiency of current AIS is also analysized. A new model named DAICA is presented for improving the AIS algorithm AISEC used for Email classification. In this model, the technique of virtual gene library was adopted to improve the updating process of the antibody in order that the information of the antibody that participated in classifying could be fully utilized. And series of experiments on UCI were made to validate the validity of the improved algorithm, meanwhile, the parameters of the algorithm were analyzed in detail. A dynamic email filtering model DTDEF based on the danger theory was also designed, and its framework along with the concrete realization arithmetic was realized. The trait of dynamic, self-adaptability and validity of classification were embodied sufficiently in filtering spam Emails.Finally, experiments on the real data we gathered from real world were performed and compared with the classical Bayes. The results show that the capability of DTDEF is better than Bayes methods' at different updating cycle. Especially when updating cycle is one, the capability of DTDEF is far better than Bayes.