一种基于人工免疫的模糊核聚类算法  

A New Artificial Immunity Based Fuzzy Kernel Clustering Algorithm

作　　者：蒋全胜[1] 贾民平[1] 胡建中[1] 许飞云[1] 

机构地区：[1]东南大学

出　　处：《中国机械工程》2008年第5期594-597,共4页China Mechanical Engineering

基　　金：国家自然科学基金资助项目(50475078)

摘　　要：针对模糊聚类及核聚类算法存在的初值敏感及易陷入局部极值点的问题,提出了一种基于人工免疫的模糊核聚类新算法。新算法将基于核的模糊聚类方法与人工免疫进化算法相结合,借鉴生物免疫系统中免疫细胞克隆和记忆、亲合力成熟等机理,采用克隆选择机制对抗体进行逐代克隆、高频变异及抑制操作。相对于模糊聚类及核聚类算法,新算法能快速地获得全局最优解。仿真数据、IRIS数据和空气压缩机运行数据测试结果证明了新算法的可行性和有效性。Aiming at the flaws of the fuzzy c-means and kernel clustering algorithm which it is sensitive to the initialization and easy to fall into local optimum, a new algorithm of fuzzy kernel clustering based on artificial immunity （AI- FKCM） was proposed. The AI- FKCM can quickly obtain the global optimum by combining kernel based fuzzy clustering and artificial immune evolution algorithm, using for reference of mechanism of immunocyte clone and memory,affinity maturation in the natural immune system, and operating to antibody with clone, high frequency mutation and restraining in each generation. The experimental results with simulation data, IRIS data and compressor running data demonstrate the feasibility and effectiveness of the new algorithm.

关 键 词：模糊聚类 人工免疫 克隆选择算法 核方法 

分 类 号：TP301.4[自动化与计算机技术—计算机系统结构]