改进的鲸鱼优化算法对牛奶脂肪含量预测模型的优化  

Optimization of Prediction Model for Milk Fat Content Based on Improved Whale Optimization Algorithm

作　　者：李欣 刘江平[1,2] 黄清 胡鹏伟 LI Xin;LIU Jiang-ping;HUANG Qing;HU Peng-wei(College of Computer and Information Engineering of the Inner Mongolia Agricultural University,Huhhot 010018,China;Inner Mongolia Autonomous Region Key Laboratory of Big Data Research and Application of Agriculture and Animal Husbandry,Huhhot 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：《光谱学与光谱分析》2023年第9期2779-2784,共6页Spectroscopy and Spectral Analysis

基　　金：内蒙古农业大学青年教师科研能力提升专项(BR 220116),国家自然科学基金项目(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)资助。

摘　　要：牛奶脂肪含量的高低会影响人的身体健康。以牛奶脂肪含量作为分析指标,应用图像处理技术分析高光谱数据,利用ENVI软件从高光谱图像中提取感兴趣区域(ROI),采用不同的预处理方法对光谱数据建立偏最小二乘回归(PLSR)模型并比较得出最佳的预处理方法,然后采用不同的主成分个数对预处理后的数据进行特征提取并建立支持向量回归机(SVR)模型,通过比较得出最佳的主成分个数,最后对特征提取后的数据建立SVR预测模型对牛奶中脂肪含量进行分析。由于传统的SVR模型预测效果不好,不能满足人们对于预测模型的基本要求,故提出一种混合策略改进的鲸鱼优化算法对SVR预测模型进行优化,将经过混合策略改进的鲸鱼优化算法优化后的SVR模型的评价参数与经过遗传算法、传统的鲸鱼优化算法、精英反向学习优化的鲸鱼优化算法优化后的SVR模型的评价参数进行了比较。结果表明:经混合策略改进的鲸鱼优化算法优化的SVR模型的训练集与预测集决定系数(R^(2))的值分别为0.998和0.995,均方根误差(RMSE)的倒数1/RMSE的值分别为13.766和6.191,平均绝对误差(MAE)的倒数1/MAE的值分别为13.910和11.422;经传统的鲸鱼优化算法优化的SVR模型的训练集与预测集参数R^(2)的值分别为0.998和0.989,1/RMSE的值分别为13.526和5.849,1/MAE的值分别为13.616和7.037;经精英反向学习策略改进的鲸鱼优化算法优化的SVR模型的训练集与预测集参数R^(2)的值分别为0.998和0.988,1/RMSE的值分别为12.474和6.421,1/MAE的值分别为15.003和10.554。由以上结果说明混合策略改进的鲸鱼优化算法优化SVR预测模型是可行的,优化后的SVR模型具有更好的预测效果。Milk fat content of high and low will affect people's health.The experiment in milk fat content analysis indicators,application of image processing technology analysis of hyperspectral data,extracting the region of interest(ROI)from hyperspectral images using the ENVI software,different preprocessing methods were used to establish Partial Least Squares Regression(PLSR)model for spectral data and the best preprocessing method was obtained by comparison,Then,different numbers of principal components were used for feature extraction of the pre-processed data and Support Vector Regression(SVR)model was established.The optimal number of principal components was obtained through comparison.Finally,the SVR prediction model was established for the data after feature extraction to analyze the fat content in milk.Since the traditional SVR model has a poor prediction effect and cannot meet people's basic requirements,this paper proposes a hybrid strategy improved whale optimization algorithm to optimize the SVR prediction model.The evaluation parameters of the SVR model optimized by hybrid strategy whale optimization algorithm are compared with those optimized by genetic algorithm,traditional whale optimization algorithm and elite reverse learning whale optimization algorithm.The results show that the training set and prediction set coefficient of determination(R^(2))of the SVR model optimized by hybrid strategy modified Whale optimization algorithm are 0.998 and 0.995,respectively.The reciprocal 1/RMSE values of Root Mean Square Error(RMSE)were 13.766 and 6.191,and the reciprocal 1/MAE values of Mean Absolute Error(MAE)were 13.910 and 11.422,respectively.The training set and prediction set parameters R^(2)of the SVR model optimized by the traditional whale optimization algorithm are 0.998 and 0.989,1/RMSE is 13.526 and 5.849,and 1/MAE is 13.616 and 7.037,respectively.The training set and prediction set parameters R^(2)of the SVR model optimized by the whale optimization algorithm improved by reverse learning strategy are 0

关 键 词：支持向量回归机 鲸鱼优化算法 黄金正弦算法 精英反向学习 模型优化 

分 类 号：O657.3[理学—分析化学]