论文部分内容阅读
目的淫羊藿作为一种药食两用植物,在祛湿补肾方面有显著疗效。应用近红外光谱,以甘肃、陕西和辽宁三个不同产区的淫羊藿为研究对象进行产地鉴别。方法采用主成分分析(principal component analysis,PCA)、前馈人工神经网络(back feed forward-artificial neural network,BP-ANN)和支持向量机(support vector machine,SVM)进行定性判别分析;其中,在支持向量机分类模型中,研究了三种参数寻优方法包括网格全局搜索(grid search)、遗传算法(genetic algorithm,GA)及粒子群算法(particle swarm optimization,PSO)对模型性能的影响。结果 PCA得分图产地间有部分重叠,较难区分;前馈人工神经网络和支持向量机定性识别方法都能完全准确地鉴别产地。结论该研究表明近红外光谱技术结合化学计量学可作为一种快速可靠的方法用于淫羊藿产地的鉴别,并为市场规范提供一种新思路。
Objective Epimedium is a kind of edible and medicinal plants, and has a significant curative effect on dampness and kidney. Near infrared spectroscopy was used to identify the origin of Epimedium in three different producing areas of Gansu, Shaanxi and Liaoning. Methods The principal component analysis (PCA), feed-forward artificial neural network (BP-ANN) and support vector machine (SVM) were used for qualitative discriminant analysis. Among them, In the support vector machine classification model, the effects of three kinds of parameter optimization methods, including grid search, genetic algorithm (GA) and particle swarm optimization (PSO), on the performance of the model are studied. Results There was some overlap between the producing areas of PCA score map, which was more difficult to differentiate. Both the feedforward artificial neural network and the support vector machine qualitative identification method could completely identify the origin. Conclusion This study shows that near-infrared spectroscopy combined with chemometrics can be used as a fast and reliable method for the identification of the origin of Epimedium and provide a new idea for market regulation.