摘要
针对单项非线性预测模型预测矿山开采地表沉陷精度低、稳定性差的问题,采用变异系数法组合互补性好的ARMA模型与BP神经网络模型,实验结果:预计值均方根误差为3. 2,平均相对误差0. 11%,结果表明组合模型在精度和稳定性方面具有较大优势。
According to the problems of low accuracy and poor stability of surface subsidence in mining by using the single nonlinear prediction model,variance coefficient method to combine complementary ARMA model and BP neural network model was used.The results showed that the estimated root mean square error was 3.2 and the average relative error was 0.11%.The combined model had the greater advantages in accuracy and stability.
作者
方苏阳
吕鑫
池深深
郭庆彪
魏涛
Fang Suyang;Lyu Xin;Chi Shenshen;Guo Qingbiao;Wei Tao(School of Geomatics,Anhui University of Science and Technology,Huainan 232001,China;JiangSu Lianwang Information Technology.Co.,Ltd.,Suqian 223800,China)
出处
《矿山测量》
2018年第6期5-9,共5页
Mine Surveying
基金
安徽省博士后基金(2014B019)
国家自然科学基金项目(41602357)
安徽高校自然科学研究项目(KJ2016A190)
关键词
ARMA
BP神经网络
变异系数法
组合预测
ARMA
BP neural network
variance coefficient method
combination forecasting