基于改进人工免疫网络的分类方法  

A Classification Method Based on Modified Artificial Immune Net

作　　者：马立玲[1] 张曌[1] 周晓航[1] 王军政[1] 

机构地区：[1]北京理工大学自动化学院,北京100081

出　　处：《北京理工大学学报》2011年第2期154-157,172,共5页Transactions of Beijing Institute of Technology

基　　金：北京理工大学基础研究基金资助项目(20080642001)

摘　　要：针对人工免疫存在的对样本要求高以及压缩阈值难以确定等问题,结合免疫算法和计算机检测的特点,提出了基于双记忆细胞层网络结构和网络邻近细胞消除规则的改进免疫方法.此外,基于物理中量子能级概率分布,提出了一种分层边界的策略.经VC环境下仿真实验,该方法在存在干扰和样本分布不理想时能够简单确定压缩阈值,有较强的适应性,提高了分类的准确率.A clustering method based on modified artificial immune is proposed to overcome the difficulties in computational effort,sampling requirements and the choice of compressing thresholds.To solve the problem on the compress parameter adjusting,a novel concept of double memories layers architecture was presented.Based on the architecture and improved cell-elimination laws,a new approach of cell-elimination process was developed.The ideas above aroused from quantum mechanics theory,the Schrdinger equation and the energy level were applied to the immune net.The simulation was performed by taking the data from UCI database.It proves that the classification accuracy of the proposed artificial immune system is improved,and better training speed is gained compared with the artifacial immune net method.

关 键 词：人工免疫 双记忆细胞层 优化消除算法 量子能级 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]