基于状态频率记忆神经网络的短时交通流预测  

Short-term Traffic Flow Prediction Based on State Frequency Memory Neural Network

作　　者：余敬柳 陈鹏[1] 谢静敏 YU Jingliu;CHEN Peng;XIE Jingmin(School of Transportation, Wuhan University of Technology, Wuhan 430063, China;Guangdong Hualu Transportation Technology Co. Ltd.,Guangzhou 510550, China)

机构地区：[1]武汉理工大学交通学院,武汉430063 [2]广东华路交通科技有限公司,广州510550

出　　处：《武汉理工大学学报（交通科学与工程版）》2020年第4期733-737,共5页Journal of Wuhan University of Technology（Transportation Science & Engineering）

基　　金：国家自然科学基金项目资助(51208400)。

摘　　要：短时交通流预测在智能交通系统中起到重要的作用.针对交通流时间序列,提出了一种基于状态频率记忆神经网络的短时交通流预测模型.该模型将交通流信息分解为状态和频率两个维度作为记忆单元进行建模,根据预测时间间隔的时长将历史交通流数据汇总,采用小波分析对历史交通流数据去噪并进行归一化处理,将其分为训练集、验证集以及测试集,最终构建短时交通流预测模型.本文以合肥市某交叉口为例,运用状态频率记忆神经网络预测该交叉口的短时交通流,并与其他预测方法的预测结果进行比较.结果表明:状态频率记忆神经网络预测短时交通流的精度更高,验证了预测模型的有效性.Short-term traffic flow prediction plays an important role in intelligent transportation system.A short-term traffic flow prediction model based on state frequency memory neural network was proposed for traffic flow time series.In this model,traffic flow information was decomposed into two dimensions:state and frequency,which were used as memory units for modeling.According to the predicted time interval,the historical traffic flow data were summarized.The historical traffic flow data was denoised and normalized by wavelet analysis,which is divided into training set,verification set and test set,and finally the short-term traffic flow prediction model was constructed.In this paper,taking an intersection in Hefei as an example,the state frequency memory neural network was used to predict the short-term traffic flow at the intersection,and the prediction results were compared with those of other prediction methods.The results show that the state frequency memory neural network has higher accuracy in predicting short-term traffic flow,which verifies the validity of the prediction model.

关 键 词：智能交通 短时交通流 状态频率记忆神经网络 预测 

分 类 号：U491.1[交通运输工程—交通运输规划与管理]