近红外光谱法最小二乘双胞胎支持向量机的应用研究

依据中药大黄的近红外光谱信息,采用最小二乘双胞胎支持向量机( LSTSVM)算法,通过MATLAB软件编程,建立参数可优化识别模型,实现了对中药大黄的真伪鉴别.将实验材料98个大黄样品随机划分为训练集和测试集,对于训练集60个样品采用留1/5法交叉验证优化模型参数,以所选最优化参数结合训练集样品的近红外光谱建立最优识别模型,对测试集的38个样品的真伪迸行识别,识别率可达97.4％.结果表明,LSTSVM算法是一种有效的识别方法,可依据中药大黄的近红外光谱对其真伪进行快速识别.同时,本研究将大黄样品6次随机划分为训练集和测试集,建模预测平均识别率为93.4％,表明采用LSTSVM算法建立识别模型具有较好的稳健性.

Application of Least Square Twins Support Vector Machine in Near Infrared Spectrometry

A method for rapid identification of the traditional Chinese herb rhubarb based on near infrared(NIR) spectroscopy was described.The identification model was established by the least square twins support vector machine(LSTSVM) algorithm with MATLAB.98 rhubarb samples were used for the investigation.To establish NIR-LSTSVM identification model,the samples were divided into training set with 60 samples and testing set with 38 samples randomly.The parameters of the model were optimized by the leave 1/5 out cross validation method for the training set.And then the optimal recognition model was established by using the selected optimal parameters and near infrared spectra.The identification rate for the testing set was 97.4%.The result indicated that it is an effective method for rapid identification of rhubarb.In addition,recognition models were established by using the above method while the rhubarb samples were randomly divided into training set and testing set six times,and the average identification rate was 93.4%.The result showed that this method presented good robustness.