基于EOBL-AO的牛奶脂肪含量预测模型  

Prediction model of milk fat content based on EOBL-AO

作　　者：刘江平[1,2] 李欣 陈晨 Liu Jiangping;Li Xin;Chen Chen(College of Computer and Information Engineering of the Inner Mongolia Agricultural University,Hohhot 010018,China;Inner Mongolia Autonomous Region Key Laboratory of Big Data Research and Application of Agriculture and Animal Hus-bandry,Hohhot 010030,China)

机构地区：[1]内蒙古农业大学计算机与信息工程学院,内蒙古呼和浩特010018 [2]内蒙古自治区农牧业大数据研究与应用重点实验室,内蒙古呼和浩特010030

出　　处：《山西农业大学学报（自然科学版）》2022年第6期40-45,共6页Journal of Shanxi Agricultural University(Natural Science Edition)

基　　金：内蒙古农业大学青年教师科研能力提升专项(BR 220116);国家自然科学基金(61962048);内蒙古科技厅关键技术攻关项目(2020GG0169);内蒙古自治区自然科学基金项目(2022MS06026);内蒙古教育厅项目(NJZY21491)。

摘　　要：[目的]牛奶中的脂肪是人所必不可少的营养物质,试验对牛奶脂肪含量进行研究,利用高光谱成像系统获取试验数据。[方法]首先,分别对经过多元散射校正(Multiplicative Scatter Correction,MSC)、变量标准化(Standard Normal Variate transform,SNV)、移动平均法(Moving Average,MA)和平滑滤波法(Savitzky-Golay,SG)四4种预处理方法得到的数据建立偏最小二乘回归(Partial Least Squares Regression,PLSR)模型,通过比较模型的评价参数得出最佳的预处理方法SNV进行接下来的操作。其次,对SNV预处理后的数据分别采取连续投影算法(successive projections algorithm,SPA)、竞争性自适应重加权算法(competitive adapative reweighted sampling,CARS)和主成分分析方法(Principal Component Analysis,PCA)进行特征选择操作并建立支持向量回归机(Support Vector Regression,SVR)模型,通过比较模型的预测精度及运行所需要的时间,得出较好的特征选择方法PCA。最后,对特征选择后的数据建立SVR模型,以达到分析牛奶中脂肪含量的目的。由于SVR模型本身的预测精度较低,本文提出一种EOBL-AO算法对SVR模型进行优化,为了检验EOBL-AO的优化效果,将该算法优化后的模型与遗传算法(Genetic Algorithm,GA)、哈里斯鹰优化算法(Harris Hawks Optimization,HHO)、天鹰座优化器(Aquila Optimizer,AO)等算法优化后的SVR模型进行比较。[结果]经EOBL-AO优化后的SVR模型的预测精度更高,且运行所需要的时间更短。[结论]利用精英反向学习(Elite Opposition-Based Learning,EOBL)改进AO是可行性的,改进后的AO算法具有更好的优化效果。[Objective]Fat in milk is an essential nutrient for human beings. This experiment was conducted to study the content of fat in milk, and the experimental data were obtained by using hyperspectral imaging system.[Methods]Firstly, the Partial Least Squares Regression(PLSR)model was established for the data obtained by Multiplicative Scatter Correction(MSC), Standard Normal Variate transform(SNV), Moving Average(MA), and Savitzky-Golay(SG)preprocessing methods. By comparing the evaluation parameters of the model, the best preprocessing method SNV was obtained for the following operations. Secondly, Successive Projections Algorithm(SPA)、Competitive Adapative Reweighted Sampling(CARS)and Principal Component Analysis(PCA)were used to select features from the pre-processed SNV data. The Support Vector Regression(SVR) model was established. By comparing the prediction accuracy of the models and the time required for operation, a better feature selection method, PCA, was obtained. Finally, the SVR model was established to analyze the fat content of milk.Due to the low prediction accuracy of the SVR model, an EOBL-AO algorithm was proposed to optimize it. To evaluate the effectiveness of the EOBL-AO optimization, the optimized SVR model was compared with those optimized by other algorithms, such as Genetic Algorithm(GA), Harris Hawks Optimization(HHO), Aquila Optimizer(AO)and other algorithms.[Results]The results indicated that the prediction accuracy of SVR model optimized by EOBL-AO was higher, and the operation time was shorter.[Conclusion]In conclusion, it is feasible to use Elite Opposition-Based Learning(EOBL)to improve AO, and the improved AO algorithm demonstrated superior optimization performance.

关 键 词：天鹰优化器 精英反向学习 主成分分析 模型优化 哈里斯鹰算法 

分 类 号：O657.3[理学—分析化学] S24[理学—化学]