摘要
大米是我国主要的粮食之一。近年来将陈大米冒充新大米售卖的现象在市场上层出不穷,严重扰乱了市场秩序。该文基于近红外光谱与机器学习方法相结合,提供了一种陈化大米以及不同程度的混掺大米的定性定量检测方法。研究中将原始近红外光谱数据采用标准正态变量变换预处理后,建立核极限学习机分别用于陈化大米的定性判别和掺假大米的定量分析,其准确度和R^(2)分别达到90%和0.8892。引入北方苍鹰优化算法用于模型的两个重要参数寻优,结果表明北方苍鹰优化算法能有效提高核极限学习机的性能,测试集预测准确度和R^(2)分别提高约5%和0.0541,为陈化大米的定性定量鉴别提供了新方法。
Rice is one of the main grains in China.In recent years,the phenomenon of passing off old rice as new rice for sale has been rampant in the upper echelons of the market,seriously disrupting market order.This article combines near-infrared spectroscopy with machine learning methods to provide a qualitative and quantitative detection method for aged rice and mixed rice to varying degrees.After preprocessing the raw near-infrared spectral data using standard normal variable transformation,a kernel limit learning machine was established for qualitative discrimination of aged rice and quantitative analysis of adulterated rice,with accuracy and R^(2)reaching 90%and 0.8892,respectively.Introducing the northern goshawk optimization algorithm for optimizing two important parameters of the model,the results show that the northern goshawk optimization algorithm can effectively improve the performance of the kernel limit learning machine,with a prediction accuracy and R^(2)improvement of about 5%and 0.0541,respectively,for the test set.This provides a new method for qualitative and quantitative identification of aged rice.
作者
倪金
索丽敏
刘海龙
NI Jin;SUO Limin;LIU Hailong(College of Information and Electrical Engineering,Heilongjiang Bayi Agricultural University,Daqing 163319,China)
出处
《食品安全导刊》
2023年第13期65-67,共3页
China Food Safety Magazine
关键词
陈化大米
掺假大米
近红外光谱
群体智能优化算法
机器学习
aged rice
adulterated rice
near-infrared spectroscopy
population intelligence optimization algorithm
machine learning
