| 用于近红外校正模型的训练样本选择 |
| |
| 引用本文： | 张其可,戴连奎.用于近红外校正模型的训练样本选择[J].传感技术学报,2006,19(4):1190-1194. |
| |
| 作者姓名： | 张其可 戴连奎 |
| |
| 作者单位： | 浙江大学智能系统与决策研究所,杭州,310027 |
| |
| 基金项目： | 国家高技术研究发展计划(863计划) |
| |
| 摘 要： | 近红外(near-infrared,NIR)校正模型建立时传统的训练样本选择方法只考虑光谱欧氏距离的同类就近选取,不考虑光谱特征异常的训练样本的影响.基于主元分析(principal component analysis,PCA)残差,在同类就近取样的基础上引入异常光谱剔除技术进行训练样本的二次提取,用于建立偏最小二乘(partial least squares,PLS)回归模型.实验结果表明,该方法比传统方法的预测精度有较明显的提高.
|
| 关 键 词： | 主元分析 近红外光谱 校正模型 训练样本选择 汽油辛烷值 |
| 文章编号： | 1004-1699（2006）04-1190-05 |
| 收稿时间： | 2005-09-22 |
| 修稿时间： | 2005年9月22日 |
Training sample selection used in NIR calibration model |
| |
| Zhang Qike,Dai Liankui.Training sample selection used in NIR calibration model[J].Journal of Transduction Technology,2006,19(4):1190-1194. |
| |
| Authors: | Zhang Qike Dai Liankui |
| |
| Affiliation: | National Key Lab of Industrial Control Technology, Institute of Intelligent Systems and Decision Making, Zhejiang University, Hangzhou 310027, China |
| |
| Abstract: | When building a near-infrared spectroscopy (NIR) calibration model, traditional training sample selection method only considers selecting nearest neighbour samples by Euclidean distance, without considering the influence of training samples with abnormal spectrum feature. Based on principal component analysis (PCA) residue, this paper introduces abnormal sample eliminating technique to selected nearest neighbour samples. These improved sample set is then applied to partial least-squares regression (PLSR) modeling. Experiment results show, introduction of abnormal sample elimination improves prediction precision of NIR model obviously. |
| |
| Keywords: | principal component analysis near-infrared spectroscopy calibration model training sample selection gasoline octane number |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
| 点击此处可从《传感技术学报》浏览原始摘要信息 |
|
点击此处可从《传感技术学报》下载全文 |
|
