Chen Yuan-Yuan, Zhang Ji-Long, Li Xiao, Tian Er-Ming, Wang Zhi-Bin, Liu Zhi-Chao
State Key Laboratory For Electronic Measurement Technology, North University of China, Taiyuan 030051, China.
Guang Pu Xue Yu Guang Pu Fen Xi. 2010 Dec;30(12):3384-7.
This paper introduced the application of support vector machines (SVM) regression method based on kernel function optimized by the rough set in the infrared spectrum quantitative calculation. According to kernel function with the rough set classification's method, the spectrum data (characteristic wavelength section) is optimized. The kernel function leads support vector machines, and the SVM project the two-dimensional room to the multi-dimensional room, and calculate the concentration of every kind of gas in multi-component pollution gas. By using two kinds of typical spectrum data processing algorithm to make the contrast, the comparison of five kinds of gaseous mixture various proximate analysis is carried out, and when the spectrum separable rate is high, the predicted values of the three methods approach the normal value, and the average error is smaller than 0.13; but when the spectrum separable rate is low, the RS-SVM predicted value is more precise than the first two kinds. Experimental data show that the consequence is better when there are more testing types, and the precision and operation of this method is of more remarkable superiority.
本文介绍了基于粗糙集优化核函数的支持向量机（SVM）回归方法在红外光谱定量计算中的应用。采用粗糙集分类方法对核函数进行处理，优化光谱数据（特征波长段）。核函数引导支持向量机，支持向量机将二维空间映射到多维空间，计算多组分污染气体中各类气体的浓度。通过与两种典型光谱数据处理算法进行对比，对五种气态混合物进行了多种近似分析比较，当光谱可分率较高时，三种方法的预测值接近正常值，平均误差小于0.13；但当光谱可分率较低时，粗糙集-支持向量机（RS-SVM）的预测值比前两种方法更精确。实验数据表明，测试种类越多效果越好，该方法的精度和运算具有更显著的优势。
