查询结果:   曹洁,沈钧珥,张红,侯亮,陈作汉.基于小波和多维重构的BP神经网络交通流短时预测[J].计算机应用与软件,2018,35(12):61 - 65,82.
中文标题
基于小波和多维重构的BP神经网络交通流短时预测
发表栏目
应用技术与研究
摘要点击数
623
英文标题
SHORT-TERM TRAFFIC FLOW FORECASTING BASED ON BP NEURAL NETWORK WITH WAVELET AND MULTIDIMENSIONAL RECONSTRUCTION
作 者
曹洁 沈钧珥 张红 侯亮 陈作汉 Cao Jie Shen Juner Zhang Hong Hou liang Chen Zuohan
作者单位
兰州理工大学计算机与通信学院 甘肃 兰州 730050 甘肃省制造业信息化工程研究中心 甘肃 兰州 730050    
英文单位
College of Computer and Communication, Lanzhou University of Technology, Lanzhou 730050, Gansu, China Gansu Manufacturing Information Engineering Research Center, Lanzhou 730050, Gansu, China    
关键词
小波降噪 相空间重构 BP神经网络 短时交通流预测
Keywords
Wavelet denoising Phase space reconstruction BP neural network Short-term traffic flow prediction
基金项目
国家自然科学基金项目(61263031);甘肃省高校科研项目(2015B-031)
作者资料
曹洁,教授,主研领域:信息融合,智能交通,智能信息处理。沈钧珥,硕士生。张红,副教授。侯亮,博士生。陈作汉,博士生。 。
文章摘要
针对采集过程中噪声影响以及交通流时间序列的强相关性,提出一种基于小波和多维重构的BP神经网络交通流短时预测方法。运用启发式小波降噪法对原始交通流数据进行降噪处理,剔除数据中的噪声;基于C-C法将交通流数据进行多维度相空间重构,充分挖掘交通流的多维变化特性;构建多维度的BP神经网络进行交通流短时预测研究。运用2 400组数据进行实验,并与传统的BP 神经网络、Elman神经网络以及SVM进行对比。实验结果表明,该方法具有较高的预测精度,绝对误差降低约2.408 0,均方误差降低约26.597 2。
Abstract
For the influence of noise in the process of traffic data collection and the strong correlation of traffic flow time series, we proposed a short-term traffic flow prediction method based on BP neural network with wavelet and multidimensional reconstruction. We adopted heuristic wavelet denoising method to reduce noise in the original data of traffic flow, and eliminated noise from the data. Based on C-C method, traffic flow data was reconstructed by multi-dimensional phase space, and the multi-dimensional change characteristics of traffic flow were fully explored. And a multi-dimensional BP neural network was constructed for the short-term prediction of traffic flow. The experiment was carried out with 2 400 sets of data and compared it with the traditional BP neural network, Elman neural network and SVM. The experimental results show that the method has higher prediction accuracy, and the absolute error decreases by about 2.408 0, and the mean square error decreases by about 26.597 2.
下载PDF全文