摘要
【目的】本研究拟通过神经网络建模来预测橡胶树白粉病病害流行趋势。【方法】采用人类专家综合利用传统经验方法预报病害的相关经验数据作为神经网络的样本,选用误差反向传播神经网络结构,以Adam算法为训练算法,成功训练得到了橡胶树白粉病神经网络预测模型。【结果】其拟合优度达88.11%,田间试验验证的实际符合率为87.88%。【结论】本研究建立的橡胶树白粉病神经网络预测模型在提供防治建议上已具备与橡胶树白粉病专家相当的水平。研究结果为进一步结合互联网通信技术和数据库管理技术,以训练完成的神经网络模型作为预测算法,建立橡胶树白粉病预测预报专家系统、实现橡胶树白粉病害预测预报流程的自动化和智能化奠定了基础。
【Objective】In recent years neural network was applied to predict diseases in many species and showed much advantages over the traditional experience methods.However little progress was reported in prediction and prediction of rubber powdery mildew based on neural network.In this study,the neural network prediction model of rubber powdery mildew was established.【Method】The quantified data from human experts system was used as training sample set,and error back propagation as neural network structure,and Adam algorithm as training method to obtain the neural network prediction model of rubber powdery mildew and further verification of the model through field experiment was performed.【Result】The goodness of fit of the neural network model established here to the test set was 99.71%,and verification of the model through field experiment showed that the actual coincidence rate was 87.88%.【Conclusion】The result indicated that the neural network prediction model of rubber powdery mildew established here had the same potentiality as human experts in providing control recommendations.Further researches combining Internet communication technology and database management technology with the trained neural network model as prediction algorithm,would help to construct an expert system to realize the automation of prediction process of rubber powdery mildew instead of human experts.
作者
叶劲秋
刘文波
林春花
郑服丛
缪卫国
YE Jin-qiu;LIU Wen-bo;LIN Chun-hua;ZHENG Fu-cong;MIAO Wei-guo(College of Plant Protection,Hainan University,Hainan Haikou 570228,China)
出处
《西南农业学报》
CSCD
北大核心
2020年第4期797-804,共8页
Southwest China Journal of Agricultural Sciences
基金
国家自然科学基金(31660033)。
关键词
橡胶树白粉病
专家系统
神经网络
病情指数
预测模型
Rubber powdery mildew
Expert system
Artificial neural network
Disease index
Prediction model
