蚁群和遗传算法优化花茶花青素近红外光谱预测模型的比较

Optimization of NIR Spectroscopy Based on Ant Colony Optimization and Genetic Algorithm for the Anthocyanin Content in Scented Tea

摘要:

以建立花茶花青素含量的最优近红外光谱模型为目标，对比研究了蚁群算法（Ant Colony Optimization, ACO）和遗传算法（Genetic Algorithm, GA）优化近红外光谱谱区的效果。ACO-iPLS将全光谱划分为12个子区间时，优选出第1、9、10共3个子区间，所建的校正集和预测集相关系数分别为0.901 3和0.864 2；交互验证均方根误差（RMSECV）和预测均方根误差（RMSEP）分别为0.160 0 mg/g和0.202 0 mg/g；GA-iPLS将全光谱划分为15个子区间时，优选出第1、5共2个子区间，所建模型的校正集和预测集相关系数分别为0.906 3和0.879 3，交互验证均方根误差（RMSECV）和预测均方根误差（RMSEP）分别为0.156 0 mg/g和0.206 0 mg/g。研究结果表明: ACO-iPLS和GA-iPLS均可以有效选择近红外光谱特征波长，其中GA-iPLS模型的精度更高。

Abstract:

Optimization of Near infrared (NIR) spectroscopy for quantitative analysis of the anthocyanin content in scented tea was discussed by selecting the optimal spectra intervals from the whole NIR spectroscopy using two variable models: Ant colony optimization interval partial least squares (ACO-iPLS) and Genetic Algorithm interval partial least squares (GA-iPLS). The ACO-iPLS full-spectrum was split into 12 intervals. The optimal intervals selected were the 1st interval, 9th interval and 10th interval. The calibration and prediction correlation coefficient of ACO-iPLS model were 0.901 3 and 0.864 2, in which the root mean square error of cross validation (RMSECV) of 0.160 0 mg/g and the root mean square error of prediction (RMSEP) of 0.206 0 mg/g were achieved. As in the GA-iPLS model, the data set was split into 15 intervals for optimization where 1st and 5th intervals were selected. The calibration and prediction correlation coefficient of GA-iPLS model were 0.901 3 and 0.864 2, and the RMSECV and RMSEP of GA-iPLS models based on these intervals were 0.156 0 mg/g and 0.206 0 mg/g, respectively. The results showed that both ACO-iPLS and GA-iPLS models could efficiently select spectrum intervals for quantitative analysis of anthocyanin in scented tea. The optimal GA-iPLS model had better performance with higher accuracy.