摘要
结合HP滤波、Elman神经网络、最小二乘支持向量机各自性质建立HP-ENN-LSSVM模型对降雨量进行预测.根据吉林省某农场1950~2015年作物生育期(5~9月)的降雨量资料,1950~2009年(5~9月)降雨量作为训练样本,2010~2015年(5~9月)的降雨量作为测试样本,验证所建模型的有效性.验证结果表明,所建模型预测效果良好,其预测平均相对误差为3.98%,与Elman、LSSVM模型相比,更适合降雨量的预测.
HP-ENN-LSSVM model is established based on HP filter,Elman neural network,and least squares support vector machine(LSSVM)and verified by the rainfall data of a farm in Jilin province during the crop growth period(May-September)between 1950 and 2015.With the rainfall data between1950 and 2009(May-September)as the training sample and the data between 2010 and 2015(May-September)as the test sample,the mean relative error of prediction is found to be 3.98%.Compared with Elman and LSSVM model,the new model ismore suitable for rainfall prediction.
出处
《玉溪师范学院学报》
2016年第4期51-56,共6页
Journal of Yuxi Normal University
基金
国家自然科学基金资助项目
编号:51066002/E060701
NSFC-云南联合基金资助项目
编号:U0937604
