XGBoost在气体红外光谱识别中的应用  

Application of XGBoost in Gas Infrared Spectral Recognition

作　　者：陶孟琪 刘家祥 吴越 宁志强[1,2] 方勇华 Tao Mengqi;Liu Jiaxiang;Wu Yue;Ning Zhiqiang;Fang Yonghua(Key Laborutory of Enrironrmental Opris and Technology,Anhui Insitute of Optics and Fine Mechanics,Chinese Academy of Sciences,Hefei,Anhui 230031,China;University of Science and Technologyof China,Hefei,Anhui 230026,China)

机构地区：[1]中国科学院安徽光学精密机械研究所环境光学与技术重点实验室,安徽合肥230031 [2]中国科学技术大学,安徽合肥230026

出　　处：《光学学报》2020年第7期195-200,共6页Acta Optica Sinica

摘　　要：为解决气体红外光谱识别问题,引入提升算法中较新的研究成果--极端梯度提升(XGBoost)算法。选用实测的三氯甲烷、对二甲苯、四氯乙烯气体的红外光谱数据进行实验。首先在对原始数据进行预处理后,通过特征工程提取光谱特征,生成特征向量;然后建立XGBoost模型,并对模型参数进行调优;最后基于分类准确率指标,将所提模型与随机森林(RF)、支持向量机(SVM)、前馈神经网络(FNN)、卷积神经网络(CNN)模型进行对比。实验结果表明,XGBoost在气体红外光谱识别领域有着广阔的应用前景。To address the problem of gas infrared spectral identification, a new lifting algorithm named eXtreme gradient boosting(XGBoost) is introduced. Infrared spectral data of chloroform, p-xylene, and tetrachloroethylene are selected for experiments. After these original data are preprocessed, the spectral features are first extracted by feature engineering to generate feature vectors. Then, the XGBoost model is established and its parameters are optimized. Finally, based on a classification accuracy index, the XGBoost model is compared with random forest(RF), support vector machine(SVM), feedforward neural network(FNN), and convolutional neural network(CNN). The experimental results show that XGBoost has a broad application prospect in the field of gas infrared spectral identification.

关 键 词：光谱学 模式识别 红外光谱 提升算法 特征工程 

分 类 号：O433[机械工程—光学工程]