二进制鼠群优化算法的特征选择及数据分类  

Feature Selection and Data Classification of Binary Rat Swarm Optimization Algorithm

作　　者：鲍美英[1] 申晋祥[1] BAO Meiying;SHEN Jinxiang(School of Computer and Network Engineering,Shanxi Datong University,Datong 037009)

机构地区：[1]山西大同大学计算机与网络工程学院,大同037009

出　　处：《计算机与数字工程》2024年第6期1612-1616,1675,共6页Computer & Digital Engineering

基　　金：国家自然科学基金项目(编号:11971277);山西省教育厅科技创新项目(编号:2019L0738);山西大同大学科研项目(编号:2020k10);山西大同大学云冈专项项目(编号:2020YGZX016);山西大同大学校级教学改革创新项目(编号:XJG2023246,XJG2023251)资助。

摘　　要：针对特征选择技术中提高分类准确率和降低特征选取个数随着数据维度增加而难度加大的问题,对新型仿生优化算法的鼠群优化算法进行改进,在算法中引入转换函数,使用K近邻法作为分类器,提出二进制鼠群优化算法,用于特征选择,进行数据分类,对特征进行有效的降维并减少数据分类的错误率。在UCI的10个数据集上进行测试,并与遗传算法、粒子群算法、樽海鞘群算法和正余弦算法进行比较,实验结果表明,所提算法能够提高数据分类准确率并有效降低特征维度,算法具有较好的收敛性和鲁棒性。In view of the difficulty of improving classification accuracy and reducing the number of feature selection in feature selection technology,with the increase of data dimension,this paper improves the new bionic optimization algorithm mouse swarm optimization algorithm,introduces the conversion function into the algorithm,uses the K-nearest neighbor method as the classifier,and proposes a binary mouse swarm optimization algorithm for feature selection and data classification,effectively reduces the di-mension of features and the error rate of data classification.It is tested on 10 data sets of UCI and compared with genetic algorithm,particle swarm optimization algorithm,bottle sea squirt swarm optimization algorithm and sine cosine algorithm.The simulation re-sults show that the algorithm can improve the accuracy of data classification and effectively reduce the feature dimension.The algo-rithm has good convergence and robustness.

关 键 词：鼠群优化算法 特征选择 数据分类 K近邻 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]