基于深度学习LSTM对交通流状态的预测

Prediction of Traffic Flow State Based on Deep Learning LSTM

当今道路交通状态对城市管理和人们出行愈加重要,影响着人类生活的方方面面.以深圳交通为研究对象,由基础车辆数据和道路坐标构建了路网系统,从车辆速度和密度两个方面导出了交通流状态评价指数TSI.利用深度学习长短期记忆神经网络(LSTM)对车辆速度和密度两个指标进行预测,并通过对比极限学习机(ELM),时间序列(ARMA)和BP神经网络,进行仿真实验,结果表明相对于传统预测模型,所采用的LSTM网络具有更优的预测精确度和对远期预测的稳定性.最后利用预测结果计算出更能直观反映出道路交通拥堵情况的TSI指数,为人们提供了准确的交通状态预测.

Nowadays,road traffic status is becoming more and more important to city management and people’s travel,which affects every aspect of human life.This paper takes ShenZhen traffic as the research object,constructs the road network system from the basic vehicle data and road coordinates,and derives the traffic flow state evaluation index TSI from the two aspects of vehicle speed and density.Predict vehicle speed and density by using the deep learning Long Short-Term Memory(LSTM).The simulation experiment is carried out by comparing the extreme learning machine(ELM),time series(ARMA)and BP neural network.The results show that compared with the traditional prediction model,the LSTM network adopted in this paper has better prediction accuracy and stability for the long-term prediction.Finally,the TSI index,which can reflect the traffic congestion more directly,is calculated by using the predicted results,providing people with accurate traffic status prediction.