基于动态人工免疫的邮件分类算法研究

Dynamic artificial immune classification algorithm for E-mail

使用虚拟基因库技术,对用于邮件分类的人工免疫系统(AISEC)进行改进,提出了动态人工免疫分类算法(DAICA)模型,改进了AISEC的抗体更新过程。当分类正确时,充分利用参与正确分类的抗体,快速改善抗体质量;当分类错误时,不再是简单地将参与错误分类的抗体直接移去,而是对这些移去的抗体进行体细胞高频变异,以保持先前遇到的抗原信息。还研究了新算法DAICA中使用的参数α与β对算法性能的影响。实验表明,这种改进可以提高分类准确率。

With the technology of virtual gene library, Artificial Immune System for Email Classification（AISEC） was improved. Model of Dynamic Artificial Immune Classification Algorithm（DAICA） was proposed and the process of updating antibody population was refined. When the classification was correct, the antibodies, which participated in the classification, should be made good use of to improve the quality of the antibodies. When the classification was incorrect, it was not to simply remove the antibodies that participated in the misclassification, but to use somatic hypermutation on these antibodies so as to reserve the information of the antigens met before. At the same time the effects of the two important parameters α and β on the performance of the algorithm DAICA were also explored. The experimental results show that this improvement can achieve higher classification accuracy.