基于REMCC-BPNN的稻瘿蚊发生量预测研究  

Study on the Forecast of Orseolia oryzae Emergency Size Based on REMCC-BPNN

作　　者：隆轲[1,2] 张红燕[1,2] 谢元瑰[1,2] 李诚[1,2] 

机构地区：[1]湖南农业大学信息科学技术学院,长沙410128 [2]湖南省农村农业信息化工程技术研究中心,长沙410128

出　　处：《中国农学通报》2014年第13期289-293,共5页Chinese Agricultural Science Bulletin

基　　金：国家科技支撑计划重大项目"农村物联网基础平台共性关键技术研究"(2012BAD35B05);湖南省研究生科研创新项目"时间序列分析方法在农业虫害预测中的应用研究"(CX2012B307)

摘　　要：为了提高预测稻瘿蚊发生量的准确度,有效防控稻瘿蚊虫害成灾面积,采用基于K近邻样本拟合相对误差绝对值与时序相关系数最小原则优化的BP神经网络预测模型REMCC-BPNN,选取广为认可的气温和降水量为影响因子,对稻瘿蚊的发生量进行独立预测。通过2个实例(化州市晚稻稻瘿蚊发生程度和广西邕宁县稻瘿蚊发生程度)验证显示:REMCC-BPNN模型的独立预测精度分别为94%和100%,明显优于经典回归分析、SVR-CAR、MIV-BPNN等参比模型。可见,REMCC-BPNN模型在虫害发生量预测方面有良好的应用前景。In order to improve the predictive accuracy of the Orseolia oryzae emergency size and control areaaffected by Orseolia oryzae effectively, an improved back-propagation neural network（BPNN） model namedREMCC-BPNN was proposed. REMCC-BPNN optimizes the training model for BPNN based on the minimumcorrelation coefficient of the absolute value of the K nearest neighbor training samples＇ fitting relative errorand the K training samples＇ time order. This study utilized air temperature and precipitation as influencingfactors to forecast pest management independently. The results of two instances（Orseolia oryzae emergencysize in Huazhou City and Yongning County of Guangxi Province） indicated that the prediction accuracy ofREMCC- BPNN was 94% and 100% respectively, which was better than that of several traditional usedmodels, such as SVR-CAR, MIV-BPNN. REMCC-BPNN had a promising application prospect in the forecastof pest management.

关 键 词：BP神经网络 虫害预测 时间序列 

分 类 号：S126[农业科学—农业基础科学]