基于自适应烟花算法和k近邻算法的特征选择算法  

FEATURE SELECTION ALGORITHM BASED ON ADAPTIVE FIREWORKS ALGORITHM AND K-NEAREST NEIGHBOR ALGORITHM

作　　者：黄欣[1] 莫海淼 赵志刚[3] Huang Xin;Mo Haimiao;Zhao Zhigang(Department of Information and Electromechanical Engineering,Guangxi Agriculture Vocational and Technical College,Nanning 530007,Guangxi,China;Research Institute of Computer Network System,School of Management,Hefei University of Technology,Hefei 230009,Anhui,China;College of Computer and Electronics Information,Guangxi University,Nanning 530004,Guangxi,China)

机构地区：[1]广西农业职业技术学院信息与机电工程系,广西南宁530007 [2]合肥工业大学管理学院计算机网络系统研究所,安徽合肥230009 [3]广西大学计算机与电子信息学院,广西南宁530004

出　　处：《计算机应用与软件》2020年第5期268-274,共7页Computer Applications and Software

基　　金：国家自然科学基金项目(61363067);广西2016年度中青年教师基础能力提升项目(KY2016YB684)。

摘　　要：特征选择是从原始特征集中选取若干个特征子集,并降低数据维度和减少冗余信息,从而达到提高分类准确度的效果。为了达到此效果,将自适应烟花算法进行离散化处理,使用k近邻算法作为分类器,并提出新的特征选择算法。将特征子集引入目标函数,并使用惩罚因子来处理约束条件,采用十折交叉验证法来检验分类效果。使用机器学习常用的UCI数据集进行仿真实验,结果表明:与增强烟花算法、烟花算法、蝙蝠算法、粒子群算法和自适应粒子群算法相比,该算法的性能更优。Feature selection is to select several feature subsets from the original feature set,and it can reduce the data dimension and redundant information,so as to improve the accuracy of classification.In order to achieve this effect,this paper discretizes the adaptive fireworks algorithm,uses k-nearest neighbor algorithm as classifier,and proposes a new feature selection algorithm.The feature subset was introduced into objective function,and penalty factor was used to deal with constraints.We used 10-fold cross validation method to test the classification effect.The UCI data sets commonly used in machine learning are used to carry out simulation experiments.Comparing with the enhanced fireworks algorithm,fireworks algorithm,bat algorithm,particle swarm optimization and adaptive particle swarm optimization,the experimental results show that the performance of our algorithm is better than the other five algorithms.

关 键 词：自适应烟花算法 特征选择 分类 K近邻算法 十折交叉验证 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]