基于神经网络的稻纵卷叶螟长期预测

A NEUTRAL NETWORK APPROACH TO LONG-TERM FORECASTING FOR RICE LEAFROLLER

为了利用神经网络强大的学习能力、非线性处理能力和预测能力,根据其建模的基本原理对江苏省通州市稻纵卷叶螟赶蛾资料进行了处理分析,建立了该地神经网络长期预报模型。结果表明:只考虑预报量的神经网络模型三年的总预测准确率达88.8％;而考虑气象因素的神经网络模型三年各项的预报准确率则高达100％。从而说明了利用神经网络模型进行害虫预测是可行的。

In this paper, a tentative long-term neutral network forecasting model for rice leafroller Gnaphalocrosis medinalis Guenee was established by using the data of systemic drive away number of adult C. medinalis from rice fields and meteorological records in Tongzhou City, Jiangsu Province as the bases of modelling. The forecast results for its occurrence time, peak and totalled moth number of 2nd and 3rd generations in three years showed that the mean accuracies of the submodels derived from moth number with and without meteorological factors were 100 and 88.8% , respectively.