摘要
近红外数据主成分降维后的因子筛选是建模的关键问题,一般是按照主成分贡献率的大小依次选择主成分因子。本文利用遗传算法进行主成分降维后的因子选择,然后利用支持向量回归方法建立烟叶中淀粉含量的定量预报模型。计算结果表明:用遗传算法和支持向量回归方法联合建模,对于训练集和独立测试集所得淀粉含量计算值与实验室之间的相关系数平方分别为0.91和0.90,结果优于利用主成分贡献率筛选变量后再用支持向量回归建模的结果。
It is very important for a model to choose suitable factors among the PCA variables available from near-infrared spectroscopy (NIRS). The subset of PCA factors is often selected based on PCA contribution rates. In this work, the genetic algorithm was applied to choose PCA factors from NIR parameters, then the support vector regress method was used to construct a model to predict the starch contents with data set of 187 tobacco samples. The correlation coefficient of square (R2) are 0.91 and 0.9 for training set and independent test set respectively. Therefore, the GA-SVR is a useful tool to develop the quantitative model for NIR parameters based on choosing PCA factors using genetic algorithm.
出处
《计算机与应用化学》
CAS
CSCD
北大核心
2014年第8期969-971,共3页
Computers and Applied Chemistry
关键词
近红外分析技术
遗传算法
支持向量机回归
near-infrared spectroscopy
genetic algorithm
support vector regression
