基于改进长鼻浣熊优化算法的牛奶脂肪光谱特征波段筛选方法研究  

Spectral Feature Band Screening Method for Milk Fat Using an Improved Coati Optimization Algorithm

作　　者：张源朴 刘江平[1,2] ZHANG Yuanpu;LIU Jiangping(College of Computer and Information Engineering,Inner Mongolia Agricultural University,Hohhot 010011,China;Inner Mongolia Autonomous Region Key Laboratory of Big Data Research and Application of Agriculture and Animal Husbandry,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,呼和浩特010011 [2]内蒙古自治区农牧业大数据研究与应用重点实验室,呼和浩特010030

出　　处：《内蒙古农业大学学报（自然科学版）》2024年第6期76-83,共8页Journal of Inner Mongolia 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项目(BR 220116);国家自然科学基金项目(61962048);国家重点研发计划项目子课题项目(2023YFD1600702-04);内蒙古自治区自然科学基金项目(2022MS06026)。

摘　　要：随着食品安全与质量控制需求的增加,近红外光谱(NIR)分析技术因其非侵入性和高效性,在快速检测领域展现出极大潜力。然而,传统特征选择方法在处理高维光谱数据时,常面临计算复杂度高、依赖性强及泛化能力不足等问题,难以满足实际应用需求。针对上述问题,本文提出了一种改进的长鼻浣熊优化算法(GD-Golden-SCOA),该算法融合了佳点集策略、动态反向学习策略和黄金正弦算法,旨在增强全局搜索能力和参数自适应性,减少人工干预。实验中,对比了5种预处理方法,并通过全波长建模、传统遗传算法(GA)、无信息变量消除算法(UVE)、原始长鼻浣熊优化算法(COA)与改进后的COA算法进行特征波段筛选。实验结果表明,改进算法不仅显著减少了特征波段数量,而且在牛奶脂肪含量预测中提高了模型的精度,训练集和测试集的R2分别达到0.994和0.989,显示出其在牛奶快速无损检测领域的广泛应用潜力。With the increasing demand for food safety and quality control,Near-Infrared(NIR)spectroscopy has shown significant potential for rapid detection due to its non-invasive and efficient characteristics.However,traditional feature selection methods often face challenges such as high computational complexity,dependency,and limited generalization when dealing with high-dimensional spectral data.To address these challenges,this paper proposed an improved Coati Optimization Algorithm(GD-Golden-SCOA)that incorporated the GoodNode strategy,Dynamic Opposite-Based Learning(DOBL),and the Golden Sine Algorithm(Golden-SA).The enhanced algorithm aimed to improve global search capability and parameter adaptability while minimizing human intervention.The study compared five preprocessing methods and evaluated the performance of full-wavelength modeling,traditional Genetic Algo⁃rithm(GA),Uninformative Variable Elimination(UVE),original Coati Optimization Algorithm(COA),and the proposed GD Golden-SCOA in selecting feature bands for predicting milk fat content.The results showed that the improved algorithm not only re⁃duced the number of selected feature bands but also significantly enhanced prediction accuracy,achieving R2 values of 0.994 for the training set and 0.989 for the test set.These findings suggested that the GD-Golden-SCOA algorithm held considerable promise for applications in rapid and non-destructive detection.

关 键 词：近红外光谱分析 特征波段筛选 长鼻浣熊优化算法 佳点集 动态反向学习 

分 类 号：O657.33[理学—分析化学]