基于群搜索优化的特征子集选择  

Selection of Feature Subset Optimized by Group Search

作　　者：谢宇 王庆龙[1] 赵春霞[2] XIE Yu;WANG Qing-long;ZHAO Chun-xia(Department of Electronic Information and Electrical Engineering at Hejfei University,Hefei Anhui 230601,China;School of Computer Science and Engineering at Nanjing University of Science and Technology,Nanjing Jiangsu 210094,China)

机构地区：[1]合肥学院电子信息与电气工程系,安徽合肥230601 [2]南京理工大学计算机科学与工程学院,江苏南京210094

出　　处：《吉林工程技术师范学院学报》2018年第8期34-36,共3页Journal of Jilin Engineering Normal University

基　　金：国家自然科学基金面上项目(61773210);国家自然科学基金面上项目(41775025)

摘　　要：在数据挖掘、机器学习、人工智能等领域中,常常需要从数据集的原始特征中选择出一些有效特征来降低数据集的维度。本文提出使用群搜索优化的方法来进行特征子集的选择。群搜索优化是一种受生物觅食行为启发的智能优化算法,该算法利用种群内部各成员之间进行的信息共享以及相互之间的分工协作来增加群体搜索到资源的概率。实验结果表明,相对于其它启发式算法,群搜索优化算法在特征子集选择中具有明显优势。It is often necessary to select certain effective features from those primitive ones of a data set to lower the dimensions in the fields of data mining,machine learning and artificial intelligence. In the paper below,group search optimization,a smart optimization algorithm inspired by biologic forage whereby a group is more likely to find out resources through information sharing and labor division among group members,was proposed in such selection. The experimental result shows that such an algorithm enjoys remarkable advantages in the selection of feature subsets compared with other heuristic methods.

关 键 词：群搜索优化 启发式算法 特征选择 降维 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]