XGBoost在气体红外光谱识别中的应用 
下载： 871次
Application of XGBoost in Gas Infrared Spectral Recognition
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
图 & 表
图 1. XGBoost算法流程
Fig. 1. Flow chart of XGBoost algorithm
下载图片 查看原文
图 2. XGBoost算法示意图[13]
Fig. 2. Schematic of XGBoost algorithm[13]
下载图片 查看原文
图 3. 光谱预处理前后对比。(a)(b)三氯甲烷;(c)(d)对二甲苯;(e)(f)四氯乙烯
Fig. 3. Comparison before and after spectral pretreatment. (a)(b) Trichloromethane; (c)(d) paraxylene; (e)(f) tetrachloroethylene
下载图片 查看原文
图 4. XGBoost模型训练流程图
Fig. 4. Flow chart of XGBoost model training
下载图片 查看原文
表 1用于气体光谱分类的特征
Table1. Features for gas spectral data classification
| Feature | Meaning |
|---|
| Width | Full width at half maximum of characteristic peak | | Kurtosis | Sharpness of characteristic peak | | Skewness | Symmetry of characteristic peak | | Correlation | Correlation coefficient with standard spectrum on NIST | | SNR | Signal to noise ratio of characteristic peak |
|
查看原文
表 2三种气体分类误差矩阵
Table2. Classification error matrix for three kinds of gases
| Gas name | Trichloromethane | Paraxylene | Tetrachloroethylene |
|---|
| Trichloromethane | 887 | 15 | 5 | | Paraxylene | 31 | 814 | 11 | | Tetrachloroethylene | 9 | 13 | 799 |
|
查看原文
表 35种模型的分类准确率
Table3. Classification accuracy for five models
| Model | Accuracy /% |
|---|
| RF | 96.35 | | SVM | 79.48 | | CNN | 80.37 | | FNN | 95.61 | | XGBoost | 96.75 |
|
查看原文
陶孟琪, 刘家祥, 吴越, 宁志强, 方勇华. XGBoost在气体红外光谱识别中的应用[J]. 光学学报, 2020, 40(7): 0730002. Mengqi Tao, Jiaxiang Liu, Yue Wu, Zhiqiang Ning, Yonghua Fang. Application of XGBoost in Gas Infrared Spectral Recognition[J]. Acta Optica Sinica, 2020, 40(7): 0730002.
PDF全文
