摘要
针对BP神经网络预测下沉系数时易陷入局部极小以及下沉系数影响因素间存在一定相关性的问题,该文提出了一种基于主成分分析(PCA)和模拟退火—粒子群优化算法(SAPSO)优化BP神经网络的下沉系数预测模型。该模型首先采用PCA对下沉系数影响因素进行降维,消除其所包含的冗余信息;然后利用SAPSO优化BP神经网络的权值与阈值;最后使用训练样本训练模型,利用训练后的模型预测5组测试样本的下沉系数,并对比分析SAPSO-BP、PSO-BP和BP神经网络模型的预测结果。实验结果表明:基于PCA-SAPSO-BP神经网络的下沉系数预测模型的预测值与实际值最为吻合,其平均绝对误差、平均绝对百分比误差及均方根误差相比SAPSO-BP、PSO-BP和BP神经网络模型显著降低,可以有效提高下沉系数预测的准确性。
Aiming at the problem that BP neural network is easy to fall into local minimum in predicting subsidence coefficient and there was a certain correlation between influencing factors of subsidence coefficient,aprediction model of subsidence coefficient was proposed based on BP neural network optimized by principle components analysis(PCA)and simulated annealing-particle swarm optimization(SAPSO).Firstly,PCA was used to reduce the dimensions of influencing factors of subsidence coefficient and eliminate the redundant information.Then the weights and thresholds of BP neural network were optimized by SAPSO.Finally,training samples were used to train the model,and subsidence coefficient of 5groups of testing samples were predicted by the trained model,the prediction results were compared and analyzed with those of SAPSO-BP neural network,PSO-BP neural network and BP neural network.The experimental results showed that the prediction values of prediction model of subsidence coefficient based on PCA-SAPSO-BP neural network were the closest to the actual values,and the mean absolute error,mean absolute percentage error and root mean square error were significantly lower than SAPSO-BP neural network,PSO-BP neural network and BP neural network,which could effectively improve the prediction accuracy of subsidence coefficient.
作者
娄高中
谭毅
白二虎
LOU Gaozhong;TAN Yi;BAI Erhu(School of Civil and Architectural Engineering,Anyang Institute of Technology,Anyang,Henan 455000,China;School of Energy Science and Engineering,Henan Polytechnic University,Jiaozuo,Henan 454000,China;State Collaborative Innovation Center of Coal Work Safety and Clean-efficiency Utilization,Jiaozuo,Henan 454000,China)
出处
《测绘科学》
CSCD
北大核心
2023年第2期124-130,147,共8页
Science of Surveying and Mapping
基金
国家自然科学基金项目(51974105,52104127)
河南省科技攻关项目(212102310406)
安阳工学院博士科研基金项目(BSJ2019028)。
关键词
主成分分析
模拟退火
粒子群优化算法
BP神经网络
下沉系数
principle components analysis
simulated annealing
particle swarm optimization
BP neural network
subsidence coefficient
