基于小生境技术和聚类分析的人工免疫算法  

Artificial Immune Algorithm Based on Niche Technology and Cluster Analysis

作　　者：郝晓丽[1] 谢克明[1] 

机构地区：[1]太原理工大学计算机学院,太原030024

出　　处：《计算机科学》2007年第9期135-138,共4页Computer Science

基　　金：国家自然科学基金(60374029);山西省回国留学人员基金(2004-18)

摘　　要：针对标准人工免疫算法存在的早熟收敛和后期收敛速度慢的问题,本文提出了一种基于小生境技术和聚类分析的改进的人工免疫算法。首先运用嵌入进化标记的小生境技术对初始种群进化,"排挤机制"有效地保持种群的多样性,防止了早熟,而标记种群的进化方向则加快了算法的收敛速度。其次聚类方法的应用使得在各极值点附近形成了聚类区域,在不同的聚类区域运用人工免疫的趋同算子和异化算子分别进行粗搜索和细搜索,以保证全局寻优的速度和精度。仿真结果表明,该改进算法较之标准免疫算法,有更快的收敛速度、更强的全局搜索能力和更好的寻优精度。Due to premature convergence and low speed of latter convergence in conventional artificial immune algo rithm, the new method is introduced in the paper which is improved artificial immune algorithm based on niche technique and cluster analysis. Firstly niche technique with revolutionary recording is taken to initial population. ＂exclusion mechanism＂ can maintain population diversity to avoid premature, and labeling evolution direction dynamically can improve convergence speed. Then cluster analysis is applied to obtain cluster areas nearby extremums. Different operators are taken in different areas respectively. Similar-taxis operator is employed to realize optimization within Cluster areas, while dissimilation operator between them. Parallel searching in coarser and finer layer can ensure the speed and precision of _global optimization. Simulation shows that the improved algorithm has higher convergence speed, better capability of-global searching and better optimization precision.

关 键 词：人工免疫算法 小生境 聚类 算子 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]