基于近红外光谱结合网格搜索-随机森林-自适应提升算法无损检测牛肉新鲜度

打开文本图片集

Abstract： To improve the prediction accuracy of beef freshness using near-infrared （NIR） spectroscopy，we proposeda predictive model based on the combination of grid search （GS），random forest （RF）and adaptive boosting （AdaBoost）. Initially，RFandAdaBoostwereemployedtostablishaNRspectroscopypredictionmodel，followedbyanalysisof the predictionaccuracyfortotal volatilebasenitrogen （TVB-N）content inbeef.Subsequently，theRFmodel，composed of multiple weak leamers，wastrained using the training set，andAdaBoost wasused to integrate these weak learners into a strong learner through varying weights to buildan ensemble model.RF was then optimized using GS to develop an AdaBoostmodelthatintegratesGS-RFas itsweak learnerforpredictingtheTVB-Ncontentinbeef.Finally，the prediction performance ofthe GS-RF-AdaBoost model based onNIR spectroscopywasanalyzedand compared with thatof the partial least square regresson，RF，AdaBoost andRF-AdaBoost models.Theresults indicatedthatthe GS-RF-AdaBoostmodel outperformed in predicting theTVB-Ncontent inbeef with the lowestroot mean square error of predicyionset and the highestcoelationcoeffcent，coefficientofeterminationandresidual predictiondeviationofpredicyionset，which were 1.731，0.969，0.924and4.331，respectively.Thesefidingsconfirm thatintegratingGS-RF-AdaBoostmodelbasedonNIR spectroscopy can effectively enhance predictive performance regarding TVB-N content in beef.

Keywords： near infrared spectroscopy;grid search;random forest;adaptive boosting;beeffreshness

DOI：10.7506/rlyj1001-8123-20250210-032

中图分类号：TS251.7 文献标志码：A 文章编号：1001-8123（2025）11-0001-08

肉类在人们的饮食中有着举足轻重的地位。（剩余17312字）