基于改进樽海鞘群算法的特征选择方法  

Feature selection method based on improved salp swarm algorithm

作　　者：申晋祥[1] 鲍美英[1] 张景安 周建慧 SHEN Jin-xiang;BAO Mei-ying;ZHANG Jing-an;ZHOU Jian-hui(School of Computer and Network Engineering,Shanxi Datong University,Datong 037009,China;Network Information Center,Shanxi Datong University,Datong 037009,China)

机构地区：[1]山西大同大学计算机与网络工程学院,山西大同037009 [2]山西大同大学网络信息中心,山西大同037009

出　　处：《计算机工程与设计》2023年第9期2714-2721,共8页Computer Engineering and Design

基　　金：国家自然科学基金项目(11971277);山西大同大学科研基金项目(2020k10);山西大同大学云冈专项基金项目(2020YGZX016);山西大同大学校级教学改革创新基金项目(XJG2021249)。

摘　　要：针对樽海鞘群算法求解精度不高、收敛速度慢、容易陷入局部最优等问题,提出融合多种策略的改进樽海鞘群算法CESSA。在初始化种群时引入无限折叠混沌映射,增加初始种群的多样性;在领导者位置更新处融入非线性收敛算子,提高算法全局勘探收敛速度,增强解的精度;根据适应度值的优劣对追随者位置分别采用收敛算子和随机数扰动方式进行更新。改进算法CESSA与K近邻分类器结合构成特征选择模型CESSAFS并用于数据分类。在8个基准函数和10个UCI数据集上分别对CESSA和CESSAFS算法进行测试,其结果表明,改进算法性能更优。Aiming at the problems of low accuracy,slow convergence and easiness to fall into local optimization of the salp swarm algorithm,an improved salp swarm algorithm CESSA was proposed with various strategies.Iterative chaotic map with infinite collapses was introduced when initializing the population to increase the diversity of the initial population.The nonlinear convergence operator was incorporated into the leader position update to improve the global exploration convergence speed of the algorithm and enhance the accuracy of the solution.According to the fitness value,the follower position was updated by convergence operator and random number disturbance respectively.The improved algorithm CESSA was combined with k-nearest neighbor classifier to form feature selection model CESSAFS,which was used for data classification.CESSA and CESSAFS algorithms were tested on 8 benchmark functions and 10 UCI data sets respectively.The results show that the improved algorithm has better performance.

关 键 词：特征选择 樽海鞘群算法 无限折叠混沌映射 全局优化 K近邻 基准函数 收敛 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]