基于文化基因算法和最小二乘支持向量机的安全数据特征处理方法  

Feature Processing Approach Based on MA-LSSVM in Safety Data

作　　者：马媛媛 施永益[2] 张宏[3] 林棋[3] 李千目[3] MA Yuan-yuan SHI Yong-yi ZHANG Hong LIN Qi LI Qian-mu(State Grid Smart Grid Research Institute, Nanjing 210003, China Zhejiang Electric Power Corporation, Hangzhou 310013 ,China School of Computer Science and Engineering, Nanjing University of Science ＆ Technology, Nanjing 210094 ,China)

机构地区：[1]全球能源互联网研究院,南京210003 [2]国网浙江省电力公司,杭州310013 [3]南京理工大学计算机科学与工程学院,南京210094

出　　处：《计算机科学》2017年第3期237-241,共5页Computer Science

基　　金：国家电网公司2015年科技项目(SGRIXTKJ[2015]216)资助

摘　　要：随着各类生物智能演化算法的日益成熟,基于演化技术及其混合算法的特征选择方法不断涌现。针对高维小样本安全数据的特征选择问题,将文化基因算法(Memetic Algorithm,MA)与最小二乘支持向量机(Least Squares Support Vector Machine,LS-SVM)进行结合,设计了一种封装式(Wrapper)特征选择方法(MA-LSSVM)。该方法利用最小二乘支持向量机易于求解的特点来构造分类器,以分类的准确率作为文化基因算法寻优过程中适应度函数的主要成分。实验表明,MA-LSSVM可以较高效地、稳定地获取对分类贡献较大的特征,降低了数据维度,提高了分类效率。With all kinds of biological intelligent evolutionary algorithms become increasingly mature, feature selection methods based on the evolutionary technology and its hybrid algorithm are emerging. According to the feature selection problem of the high dimensional small sample safety data, this paper combined memetic algorithm （MA） and least squares support vector machines （IAS-SVM） to design a kind of wrapper feature selection method （MA-LSSVM）. The proposed method utilizes the specialty of least squares support vector machine being easy to search optimal solution to construct classifier, then regards classification accuracy as the main component of memetic algorithm fitness function in the optimization process. The experimental results demonstrate that MA-LSSVM can be more efficient and stable to ob- tain features larger contribution to the classification precision, and can reduce the data dimension and improve the classi- fication efficiency.

关 键 词：特征选择 文化基因算法 最小二乘支持向量机 稳定性 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]