摘要
为实现对山东烟区有翅蚜迁飞高峰期的预测预报,将有翅蚜迁飞高峰期划分为4级,以10年的历史资料为基础数据,采用逐步回归以及BP神经网络2种方法分别建立了有翅蚜迁飞高峰期预测模型。这2种方法对待测样本的预测准确度分别为95.00%、97.67%,回测准确度分别为96.65%、98.74%。所建立的预测模型可提前1个多月对有翅蚜迁飞高峰期进行预测,为中期预测模型,其预测结果可为烟田蚜虫及蚜传病毒病的防治提供依据。
Two prediction models were developed to predict occurrence tendency of alate aphids migration. The peak period of alate aphids migration was divided to 4 levels in Shandong tobacco growing region. Based on data in the past 10 years, stepwise regression and BP neural network were employed to establish prediction models of alate aphids migration peak period, respectively. The prediction accuracy of the two methods reached 95.00 % and 97.67 %, and back prediction accuracy reached 96.65 % and 98.74 %, respectively. The alate aphids migration peak period can be predicted one month in advance through the established prediction model, which can serve as a middle-term prediction model applied in the control of tobacco aphid and aphid-borned tobacco virus diseases.
出处
《中国烟草学报》
EI
CAS
CSCD
2009年第2期71-74,79,共5页
Acta Tabacaria Sinica
基金
山东省烟草专卖局科技项目(KN90)
关键词
有翅蚜
逐步回归
BP神经网络
预测模型
alate aphids
stepwise regression
BP neural network
prediction model