论文部分内容阅读
为了提高近红外光谱快速检测烟草尼古丁含量的精度和稳定性,利用近红外光谱结合遗传算法-最小二乘支持向量回归(GA-LSSVR)建立了回归预测模型。在LSSVR模型建立过程中,采用遗传算法对LSSVR参数进行自动优化。相比于利用常规最小二乘支持向量机和遗传偏最小二乘法等建立的回归预测模型,GA-LSSVR法建立的回归预测模型泛化能力更强,预测效果更好,验证集相关系数R2为0.9766,预测均方根误差为0.1065。研究结果表明,GA-LSSVR是一种快速准确的建模方法,为烟草尼古丁含量的近红外测定和近红外光谱数据的处理提供了新的方法与途径。
In order to improve the accuracy and stability of nicotine content in tobacco by near infrared spectroscopy, a regression prediction model was established by using near-infrared spectroscopy and genetic algorithm - least squares support vector regression (GA-LSSVR). In the process of establishing LSSVR model, genetic algorithm is used to optimize LSSVR parameters automatically. Compared with the regression prediction model established by the conventional least squares support vector machine and genetic partial least squares, the regression prediction model established by the GA-LSSVR method has better generalization ability and better prediction effect, and the correlation coefficient R2 of the validation set is 0.9766, the root mean square error of prediction is 0.1065. The results show that GA-LSSVR is a fast and accurate modeling method, providing a new method and approach for the near-infrared determination of nicotine content and near-infrared spectroscopy data processing.