基于选择性集成学习的高速列车故障识别研究  

Study on recognition of high speed rail malfunction based on selective ensemble learning

作　　者：饶川 苟先太[1] 金炜东[1] Rao Chuan;Gou Xiantai;Jin Weidong(School of Electrical Engineering,Southwest Jiaotong University,Chengdu 610031,China)

机构地区：[1]西南交通大学电气工程学院,成都610031

出　　处：《计算机应用研究》2018年第5期1365-1367,共3页Application Research of Computers

基　　金：国家自然科学基金重点资助项目(61134002);国家自然科学基金资助项目(61075104);中央高校基本科研业务费专项资金资助项目(SWJTU11BR039;SWJTU11ZT06);四川省科技计划项目-重点研发项目(2017GZ0159)

摘　　要：在SVM分类识别中,分类器模型一经训练得到,对所有测试样本进行无差别的识别。针对高速列车故障中样本的分类识别是存在区域分类精度的情况,提出了一种基于选择性集成学习的SVM多分类器融合算法。该方法选取测试样本最邻近的k个训练样本;然后选择对其分类效果好的SVM分类器进行融合,以提高分类准确率;最后使用高速列车故障数据进行了实验,并与Ada Boost、KNN、Bayes、SVM分类方法进行了比较。实验结果表明,该算法提高了分类识别准确率。In the classification and recognition of SVM,classifier model will identify all the testing samples without difference after it’s formed.Regarding that accuracy grade of region classification exists in classification and recognition of high-speed rail malfunction samples,this paper proposed a multi-classifier fusion algorithm based on selective ensemble learning.The process of this algorithm first selected the nearest training samples by testing samples of k numbers and then fused them with effective SVM classifiers to increase accuracy rate.And it conducted an experiment with high-speed rail malfunction data,compared this algorithm with AdaBoost,KNN,Bayes and SVM classifications.The experimental results show that the algorithm improves accuracy rate of classification and recognition.

关 键 词：选择性集成学习 支持向量机 多分类器融合 区域分类精度 高速列车故障分类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]