一种人工蜂群算法优化的邻域粗糙集特征选择方法  

Attribute Selection Method Based on Artificial Bee Colony Algorithm and Neighborhood Resolution Matrix Optimization

作　　者：季雨瑄 叶军[1,2] 杨震宇 敖家欣 JI Yuxuan;YE Jun;YANG Zhenyu;AO Jiaxin(College of Information Engineering,Nanchang Institute of Technology,Nanchang 330000,China;Jiangxi Province Key Laboratory of Water Information Cooperative Sensing and Intelligent Processing,Nanchang 330000,China)

机构地区：[1]南昌工程学院信息工程学院,江西南昌330000 [2]江西省水信息协同感知与智能处理重点实验室,江西南昌330000

出　　处：《郑州大学学报（理学版）》2023年第6期55-62,70,共9页Journal of Zhengzhou University:Natural Science Edition

基　　金：国家自然科学基金项目(61562061);江西省教育厅科技项目(GJJ211920,GJJ170995)。

摘　　要：在邻域粗糙集模型中,由于计算邻域及正域的工作量较大,因此,对邻域决策表进行特征选择或降维具有较高的时间复杂度。特别是随着邻域决策表维数的增多,计算工作量呈指数级增加。针对此类问题,引入人工蜂群算法进行优化。首先,给出了一种邻域粗糙集分辨矩阵特征重要性度量方法;其次,以邻域分辨矩阵特征重要度为启发因子构造了适应度函数,新的适应度函数增加了启发信息;最后,设计了一种人工蜂群算法优化的特征选择算法。UCI数据集对比实验结果与分析表明,与原有的邻域特征选择算法相比,新算法减少了迭代次数,加快了收敛速度,并且能够有效寻找到最小特征子集。In the neighborhood rough set model,due to the large workload of computing neighborhoods and positive regions,the feature selection or dimensionality reduction of neighborhood decision tables has high time complexity.Especially as the dimension of the neighborhood decision table increases,the computational workload increased exponentially.For such problems,artificial swarm algorithm was introduced for optimization.Firstly,a method for measuring the importance of neighborhood rough set discrimination matrix attributes was given.Secondly,a fitness function was constructed with the attribute importance of neighborhood discrimination matrix as the heuristic factor,and the new fitness function added heuristic information.Finally,A feature selection algorithm optimized by artificial bee colony algorithm was designed.The experimental results of UCI dataset comparison showed that,compared with the original neighborhood feature selection algorithm,the new algorithm reduced the number of generations,accelerated the convergence speed,and could effectively find the minimum feature subset.

关 键 词：邻域决策系统 分辨矩阵 人工蜂群算法 特征选择 特征重要度 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]