摘要
为探索基于常规监测数据的神经网络预警模型在农产品传统风险管理中的应用,以2011—2012年我国5省市的蔬菜中农药残留监测数据为样本,采取神经网络方法建立风险预警模型。首先,以产品种类、监测环节、监测时间和蔬菜产地为参考采用专家打分法将样本进行安全性评级,然后将经过筛选和预处理的45种农药监测数据,作为BP神经网络输入层,并根据不安全蔬菜的风险程度,以非常安全(A)、比较安全(B)、基本安全(C)、较不安全(D)和不安全(E)5个等级作为输出层,农药残留数据经过处理整合后得到16个样本,通过对其中14个样本进行拟合训练,得到预警模型及2个验证样本的评分结果分别为2.343 0和3.171 5,与实际评分结果隶属同一安全等级。证明基于客观监测数据的神经网络预警模型对于蔬菜中农药残留的预警是有效的。
In order to explore the application of ANN early warning model,based on routine monitoring data in the traditional risk management of agricultural products,the pesticide residues in vegetable monitoring data from China＇s five provinces from 2011 to 2012were taken as samples and the ANN risk early warning model was established.First,the experts rated the safety of samples,taking product variety,monitoring link and monitoring time and vegetable origin for reference.Then 45 kinds of pesticides filtered and monitoring data preprocessed were taken as the BP neural network input layer.And the five degrees of very safe（A）,relatively safe（B）,basic safe（C）,relatively unsafe（D）and unsafe（E）were set as the output layer.Choosing 14 samples as training samples of the ANN warning model,the results of 2validation samples were 2.3430 and 3.171 5 respectively.The model results and actual results belong to the same safety level.It proved that the ANN early warning model based on the objective monitoring data is valid for the pesticide residue risk in vegetables.
出处
《中国农业大学学报》
CAS
CSCD
北大核心
2015年第2期259-267,共9页
Journal of China Agricultural University
基金
中国农业科学院基本科研业务费项目支持(2012ZL016)
关键词
BP神经网络
预警模型
蔬菜农药残留
监测数据
BP neural network
early-warning model
pesticide residues of vegetables
monitoring data
作者简介
张星联，助理研究员，博士，主要从事农产品质量安全风险评估和预警研究，E-mail：zxlcaas@163．com
通讯作者：唐晓纯，教授，主要从事食品安全预警研究，E-mail：tangxc@rue．edu．cn
