基于多种模型组合的短时交通流预测  

Short-term Traffic Flow Prediction Based on a Combination of Multiple Models

作　　者：王德广[1] 张轶楠 WANG Deguang;ZHANG Yi’nan(College of Environmental Science and Engineering,Dalian Jiaotong University,Dalian 116028,China)

机构地区：[1]大连交通大学,软件学院,辽宁大连116028

出　　处：《微型电脑应用》2022年第3期130-133,共4页Microcomputer Applications

摘　　要：日益增加的交通流量使得道路交通面临着严峻考验,运用现代技术手段,对短时交通流进行精准的预测,能为改善及便利道路交通管理。为此提出一种基于长短时记忆神经网络(LSTM)、门控循环单元(GRU)、栈式自编码器(SAE)以及简单循环单元(SRU)模型相结合的短时交通流预测模型——LGSS模型。实验表明LGSS组合模型的预测效果,从多个评价指标分析,相较于传统的单一模型都有较大改善。同时利用SRU可进行并行运算的特点,当迭代4次预测准确度最高时,单次计算时间达到0.7491 s,比传统模型减少了约30%的时间。The increasing traffic flow makes road traffic face on a severe test.Using modern technology to accurately predict short-term traffic flow will bring convenience to improve and manage the road traffic.A long-and short-term traffic flow prediction model—LGSS combination model based on the combination of LSTM,GRU,SAE and SRU models is proposed.Experiments show that the predictive effect of the LGSS combination model is improved after the analysis of multiple evaluation indicators compared with the traditional single model.At the same time,the SRU can be used to carry out parallel computing.When the prediction accuracy of four iterations is the highest,the single calculation time reaches 0.7491 s,which is about 30%less than that of the traditional model.

关 键 词：短时交通流预测 神经网络模型 深度学习 组合预测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]