摘要
使用陕西长武1984～1998年的烟草病情资料及当地气象资料作为样本,将遗传算法与反向传播网络(BP网络)应用于烟草病毒病预测。在隐含层间使用变形的sigmoid函数,利用遗传算法优化网络权值和参数。训练时,样本中加入随机噪声,使用表决网综合输出。同时,改进了传统BP网络的收敛速度和泛化能力,预测结果基本符合实际值。
On the basis of genetic algorithm and back propagation network, the state of tobacco virus diseases was forecasted and predicted in this paper, according to the diseases state and climate data of Changwu city, Shanxi province from 1984 to 1998. The distorted sigmoid function was used as the active function in the hidden layer. The net power value and parameters were optimized by genetic algorithm. The random noise was added in training sample and exported by voted net. Constringency velocity and generalization ability was improved by amending traditional BP net. The value was tallied with forecast result.
出处
《农机化研究》
北大核心
2008年第3期66-69,共4页
Journal of Agricultural Mechanization Research
关键词
植物保护
烟草病毒病
理论研究
BP网络
遗传算法
随机噪声
plant protection
tobacco virus disease
theoretical research
BP network
genetic algorithm
random noise
