摘要
用神经网络测报系统软件,以山东郓城棉铃虫三代卵量、与山东郓城棉铃虫三代卵量显著相关的北太平洋海域前期海温距平平均值数据为例,对初始权值、隐含层神经元个数、偏差、学习系数等参数的变化对经学习训练后模型收敛的局部误差极小值点的变化和模型预测效果的影响进行了详细的探讨。
Taking the data of egg counts of the third generation cotton bollworm（CBW） in Yuncheng, Shandong Province, and the data of early mean Sea Surface Temperature（SST）anomaly in the North Pacific Ocean significantly correlative with the egg counts of the third generation CBW in Yuncheng as examples, the influence of different parameters, such as the initial weights, the number of neurons in hidden layer, the deviation and the learning coefficient, on the variation of the local error minimum point and the forecast effect of the convergent neural network model after learning and training were explored with neural network forecast system.
作者
张晓红
王晓霞
徐明
苏万力
ZHANG Xiaohong;WANG Xiaoxia;XU Ming;SU Wanli(College of Science and Information,Qingdao Agricultural University,Qingdao 266019,China)
出处
《青岛农业大学学报（自然科学版）》
2018年第4期308-315,共8页
Journal of Qingdao Agricultural University(Natural Science)
关键词
神经网络模型
病虫预报
预测测报
预警系统
棉铃虫
artificial neural network
prediction of pests
forecasting
early warning system
cotton bollworm
