基于混沌鲶鱼效应的人工蜂群算法及应用

Artificial Bee Colony Algorithm with Chaotic Catfish Effect and Its Application

针对目前人工蜂群算法的早熟收敛、陷入局部极值等问题,提出一种基于混沌鲶鱼效应的改进人工蜂群算法.首先,采用随机性更高的混沌序列初始化蜂群以扩大其遍布范围;其次,集成了鲶鱼效应和混沌理论提出了混沌鲶鱼蜂,并引入了它与跌入局部极值的蜂群之间的有效竞争协调机制,从而增进蜜蜂群体跳出局部最优解、加速收敛的能力.支持向量机的学习能力主要取决于其惩罚因子C和核函数参数的合理选择,对其参数的优化可以提升其学习效果,然而现行算法均存在一定局限性.基于我们提出的改进人工蜂群算法,对支持向量机的参数进行了优化.最后,在UCI(加州大学欧文分校)数据集和行为识别真实数据集上进行了测试,验证基于改进人工蜂群算法的支持向量机具有更强的分类性能.

There are the disadvantages of easily falling into premature convergence and local optimal solution which the ele-mentary artificial bee colony algorithm had in some degree .Chaotic Catfish effect was hence adopted in this paper to achieve the op-timum performance of artificial bee colony algorithm ,in which ,chaotic mechanism was conducted to instantiate each individual of the swarm firstly owing to its marvelous intrinsic randomness .Then the efficacious competition and coordination mechanism among Catfish bees which were derived from the integration of Chaos theory with Catfish effect and originals were intended to boost the ca-pabilities of them leaping out of local optimal solution and converging expeditiously .The practicability of Support Vector Machines （SVM ）is excessively affected due to the difficulty of selecting appropriate penalty factor C and kernel function parameter of SVM . Conversely ,all of the common SVM parameters optimization methods have their respective disadvantages with some degree of com-petence .We utilized the improved artificial bee colony algorithm to optimize the two parameters of SVM ,simultaneously ,the public datasets from the University of California-Irvine （UCI ）and the activity recognition reality data were employed for evaluating the pro-posed model .Experimental results demonstrate that the classification accuracy obtained by the developed SVM was higher.