人工免疫C_均值聚类算法.pdfVIP

39 8 西 安 交 通 大 学 学 报 Vol. 39 8 2005 8 JOU RNAL OF XIAN JIAOT ONG U NIVERSIT Y Aug . 2005 C 1, 2 1 张 雷 , 李人厚 ( 1. , 7100 9, ; 2. , 71003, ) : 通过借鉴生物免疫系统中的克隆选择原理和记 机制, 提出了一种人工免疫 C均值混合聚类算法. 该算法采用了新的克隆选择方法, 通过亲和度排序和 个体浓度定义了 个体的选择概率, 从而可确定 个体的适 应值评价函数, 以评价和选择 个体. 算法还集成了一种 C均值搜索算子, 用于加快收敛速度. 在聚类数目已 知的情况下, 所提算法能够得到给定数据集下的全局最优划分, 与基于遗传算法的聚类方法比较, 它具有更 快的收敛速度和更高的收敛精度, 并可扩展到性能指标能够表示为优化聚类中心函数的聚类模型之中. 仿真 结果表明, 所提算法是有效性的. : 聚类算法; 人工免疫; C均值 : TP183 : A : 0253987X( 2005) 0808360 Artificial Immune CMeans Clustering Al gori thm Zhang L ei 1, 2, L i R enhou 1 ( 1. Institute of System Engineering, Xian Jiaotong University, Xian 7100 9, China; 2. School of Electronics and Inform ation Engineering, H enan University of Science and Technology, Luoyang 71003, China) Abstract: Inspired by the clone selection principle and memory mechanism of the vertebrate immune sys tem, a hybrid algorithm combining Cmeans algorithm and artificial immune algorithm is presented. A new clone selection strategy is used and the individual selection probability is defined through sorting the affini ty and individuals concentration so that the evaluating function of the individual fitness can be determined, and then individuals are evaluated and selected. The Cmeans algorithm is treated as a new search operator in order to improve the convergence speed. Comparing w ith the genetic algorithm based clustering approa ches the proposed algorithm can converge to the global optimum faster and has higher accuracy. Given the cluster number, the algorithm can obtain the best partition of data sets. T he algorithm can be extended to other clustering model w hose objective function can be represented in terms of op