论文部分内容阅读
为实现鸭蛋蛋清中庆大霉素(GM)残留含量的快速测定与检测模型精度的提高,应用遗传算法(GA)筛选导数同步荧光光谱特征波长,用遗传-支持向量回归(GA-SVR)建立鸭蛋蛋清中GM残留含量的预测模型。首先分析了样本的三维同步荧光光谱和确定了本实验研究的波长差Δλ为120nm;然后利用sym5小波的2层分解对一阶导数同步荧光光谱进行去噪处理,并利用GA筛选出了14个荧光特征波长;最后利用GA优化了SVR的径向基核函数(RBF)参数(c,g,p),进而比较了GA-SVR、PLS和MLR 3种预测模型的预测能力,研究表明,以GA-SVR模型的预测能力最强,其预测集的决定系数(R2)和均方根误差(RMSEP)分别为0.983 0和1.149 4mg/L。实验结果表明,GA能有效筛选出鸭蛋蛋清中GM的荧光特征波长和提高GA-SVR模型预测精度。
In order to achieve the rapid determination of gentamicin (GM) residues in duck egg white and the improvement of the detection model accuracy, genetic algorithm (GA) was used to screen the characteristic wavelength of derivative synchronous fluorescence spectrum and establish genetic-support vector regression (GA-SVR) Prediction Model of GM Residues in Duck Egg. Firstly, the three-dimensional synchronous fluorescence spectrum of the sample was analyzed and the wavelength difference Δλ of the experimental study was determined to be 120 nm. Then, the first derivative synchronous fluorescence spectrum was denoised by the second decomposition of sym5 wavelet, and 14 Finally, the Radial Basis Function (RBF) parameters (c, g, p) of SVR were optimized by using GA, and the predictive ability of GA-SVR, PLS and MLR prediction models was compared. The prediction ability of GA-SVR model is the strongest, and its determination coefficient (R2) and root mean square error (RMSEP) of prediction set are 0.983 0 and 1.149 4 mg / L, respectively. The experimental results show that GA can effectively screen out the fluorescent characteristic wavelength of GM in duck egg white and improve the prediction accuracy of GA-SVR model.