台州市马尾松毛虫发生趋势预测技术探讨

Forecast Techniques on Occuring Current of Pine Caterpillar in TaiZhou City,Zhejiang province

运用人工神经网络的原理和方法,根据相关系数法选取与台州市7个县(市、区)马尾松毛虫有虫面积相关关系都比较密切的14个气象因子,然后进行主成分分析,在此基础上,将前6个主成分的主坐标值作为样本的输入特征,建立7个县(市、区)马尾松毛虫有虫面积分类预测的BP网络模型,结果表明:所建立的BP模型,具有令人满意的拟合精度和预测精度。当隐含层神经元个数为8个时,11年共77组用于马尾松毛虫有虫面积预测建模样本的拟合率为85.71%,7个县(市、区)3年共21组预留样本检验报准率为80.95%。

According as correlative coefficient, select thirteen weather genes with better correlativity as the pine caterpillar's pest areas in seven TaiZhou's counties (cities or towns) by handling principle and ways of manpower's NN. And then, analyzing primary elements. On the baise, set up a pine caterpillar pest sort forecast BP network modle in seven TaiZhou's counties (cities or towns) by importing characters with six former element coordinate cost stylebooks. The result showed that the BP modle is better famous simulation forecast precision. When the concealed Nerve cell's number is eight, eleven year's combining as heigh as 85.71%,with seventy-seven stylebooks of pine caterpillar pest areas,three's checking up rate of twenty-one obligated stylebook is as heigh as 80.95%.