摘要
针对交通拥堵严重影响人民日常生活,制约社会经济发展的问题,提出了基于轨迹大数据的交通拥堵评估和预测方法,旨在为相关部门提供决策指导。在交通拥堵评估中,使用交通流参数计算道路的交通状况综合参数C;在交通拥堵预测中,使用深度学习技术建立交通拥堵预测模型,将长短期记忆模型(LSTM)与向量回归模型(SVR)和循环神经网络模型(RNN)进行对比,最后发现LSTM模型的预测效果最好。
Aiming at the problem that traffic congestion seriously affects people’s daily life and restricts social and economic development,this paper proposes a traffic congestion assessment and prediction method based on trajectory big data,which aims to provide decision-making guidance for relevant departments. In the traffic congestion assessment,traffic flow parameters were adopted to calculate the comprehensive parameter C of road traffic conditions. In the traffic congestion prediction,deep learning technology was utilized to build a traffic congestion prediction model. The Long Short-Term Memory( LSTM) was compared with the Support Vector Regression model( SVR) and the Recurrent Neural Network model( RNN) for investigation. As a conclusion,the LSTM model is found to have the premium prediction results.
作者
温美玲
路鹏远
蔡林
程洋溢
WEN Meiling;LU Pengyuan;CAI Lin;CHENG Yangyi(School of Geodesy and Geomatics,Wuhan University,Wuhan 430070,China;School of Computer Science,Wuhan University,Wuhan 430070,China)
基金
国家大学生创新训练计划资助项目(201810486092)
关键词
轨迹大数据
交通拥堵
评估与预测
深度学习
长短期记忆模型
trajectory big data
traffic jam
assessment and prediction
deep learning
long short-term memory(LSTM)
