BP神经网络在烟蚜发生程度预测中的应用  

Application of BP Neural Network to Predict Occurrence Degree of Myzus persicae

作　　者：任广伟[1] 王秀芳[1] 王新伟[1] 李晓 张建党 梁其涛 

机构地区：[1]中国农业科学院烟草研究所,山东青岛266101 [2]山东中烟工业公司技术中心,山东济南250013 [3]安康市烟草公司,陕西安康725000

出　　处：《华北农学报》2008年第B10期373-376,共4页Acta Agriculturae Boreali-Sinica

基　　金：山东省烟草专卖局科技项目(KN90)

摘　　要：为实现对烟田烟蚜发生程度的预测预报,以12年的历史资料为基础数据,采用BP神经网络方法建立了烟蚜发生程度的预测模型。该模型对待测样本的预测准确度为99.43%,回测准确度为87.36%。所建立的预测模型可提前1个多月对烟蚜发生程度进行预测,为中期预测模型,其预测结果可为烟田蚜虫综合治理提供依据。Basing on the data of 12 years,the method of BP neural network was applied to establish the prediction model of Myzus persicae occurrence for its prediction and forecast in tobacco field.The model exhibited prediction accuracy of 99.43%,and back prediction accuracy of 87.36%,respectively.The prediction model could predict the occcurence degree of Myzus persicae more than one month in advance,so the middle-term prediction model could be applied in IPM of Myzus persicae in tobacco field.

关 键 词：烟蚜 发生程度 BP神经网络 预测模型 

分 类 号：S435.72[农业科学—农业昆虫与害虫防治]