基于LM—BP网络粮食产量预测.docVIP

基于LM—BP网络粮食产量预测摘要：利用Levenberg-Marquardt（LM）算法对BP神经网络法进行改进，提出了基于改进型LM-BP神经网络模型的粮食产量预测方法。提取了粮食作物播种面积、化肥施用量、粮食作物有效灌溉面积、受灾面积、农村用电量、农业机械总动力、从事农业的人口、农村居民家庭生产性固定资产原值、农村居民家庭平均纯收入9个因子作为输入因子构筑模型，粮食产量作为网络输出，通过LM算法使网络误差最小化，最后使用相关系数、相对误差等指标对模型的模拟结果进行检验。结果表明，训练样本集中模拟值和实际值的相关系数为0.996，平均相对误差为0.47%；检测样本集中，预测值和实际值的相关系数为0.994，平均相对误差为0.56%；该模型具有较高的拟合精度和预测精度，将此网络模型应用于粮食产量预测是有效的、可行的。 关键词：LM-BP网络；粮食产量；预测 中图分类号：S11+4；TP183 文献标识码：A 文章编号：0439-8114（2012）23-5479-03 Forecasting Corn Production Based on LM-BP Neural Network GUO Qing-chun1，3，4，HE Zhen-fang2，4，LI Li3 （1. Teaching Affairs Office， Shaanxi Radio TV University， Xian 710068， China； 2. Cold and Arid Regions Environmental and Engineering Research Institute， Chinese Academy of Sciences， Lanzhou 730000， China； 3. Institute of Earth Environment Research， Chinese Academy of Sciences， Xian 710075， China； 4. Graduate University， Chinese Academy of Sciences， Beijing 100049， China） Abstract： A corn production porecasting method based on improved LM-BP was proposed. According to measurement and agricultural significance principle， 9 factors of grain-sown area， fertilizer input， effective grain irrigated area， stricken area， rural electricity consumption， total agriculture mechanism power， the population engaged in agriculture， rural residents family productive assets， the average net income of rural households were extracted as the network input； corn production was extracted as the network output. The LM algorithm could minimize the error， and the modeling results were evaluated with the correlation coefficients， relative error， etc. For training sample set， the correlation coefficient between the simulated value and the actual value was 0.996， the average relative error was 0.47%； for testing sample set， the correlation coefficient between the forecasted value and the actual value was 0.994， the average relative error was 0.56%. The results showed that the improved LM-BP model could improve simulation precision and stability of the model. This