融合蝠鲼旋风式觅食策略的灰狼特征选择算法  

A Grey Wolf Feature Selection Algorithm Incorporating a Whirlwind Foraging Strategy for Manta Rays

作　　者：杨舒云 刘宏志[1] 李海生[1] YANG Shu-yun;LIU Hong-zhi;LI Hai-sheng(School of Computer Science and Technology,Beijing Technology and Business University,Beijing 100048,China)

机构地区：[1]北京工商大学计算机学院,北京100048

出　　处：《计算机仿真》2023年第9期375-380,共6页Computer Simulation

基　　金：国家自然科学基金资助项目(61877002);北京市自然科学基金-丰台轨道交通前沿研究联合基金资助项目(L191009);北京市教委-市自然基金委联合资助项目(KZ202110011017)。

摘　　要：针对传统灰狼算法在进行特性选择时易陷入局部最优,求解精度不高等问题,提出融合蝠鲼旋风式觅食策略的混合灰狼算法(BWSGWO)。在初始阶段采用混沌映射,降低初始分布的随机性,提高种群多样性;在捕猎阶段引入蝠鲼旋风式螺旋觅食的新型捕猎行为,协调算法的勘探与开采能力;最后基于灰狼优化后期个体均向领导区域聚集的问题,对最优个体采用自适应精英反向学习机制,以提升抗局部极值能力。使用UCI的8个规模各异的数据集与几种经典智能算法进行对比实验,结果表明,在相同的参数设置下,所提算法可以取得更高的分类精度和更少的特征数。In order to solve the problem that the traditional gray wolf algorithm is prone to fall into local optimization when selecting characteristics,and its solution accuracy is not high,a hybrid gray wolf algorithm(BWSGWO)integrating the manta ray whirlwind feeding strategy is proposed.In the initial stage,a chaotic mapping is used to reduce the randomness of the initial distribution and improve the population diversity;in the hunting stage,a new hunting behaviour of manta cyclone spiral foraging is introduced to coordinate the exploration and exploitation ability of the algorithm;finally,based on the problem that all individuals in the late stage of grey wolf optimisation converge to the leader region,an adaptive elite reverse learning mechanism is used for the optimal individuals to improve the resistance to local extremes.Comparative experiments using eight datasets of varying sizes from the UCI with several classical intelligent algorithms show that the proposed algorithm can achieve higher classification accuracy and a lower number of features with the same parameter settings.

关 键 词：特征选择 混沌映射 灰狼优化 旋风式螺旋觅食 精英反向学习 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]