基于多元时间序列预测的智能交通系统
|
摘要
随着城市化进程的高速发展,交通拥堵已成为困扰和阻碍城市发展的重要问题。道路大多数是部分拥堵、部分畅通,准确预测出道路拥堵状态可以更好地实现汽车分流,缓解交通压力。本文分别运用VARMA(向量自回归移动平均)和LSTM(长短期记忆网络)算法对首都机场附近的57条道路的拥堵数据进行建模分析,在此基础上将LSTM处理多元时间序列的核心思想加入到多元回归算法中,使多元回归算法拥有处理多元时间序列的能力。之后对三个算法的预测准确度和建模复杂度进行对比,找出适合用于不同场景的算法。得出结论,VARMA模型适用于短期精准预测、RNN适用于长期大规模的波动预测、改造后的多元回归模型适用于中长期快速预测。本文中的算法和结论可以更好地帮助公安和交警及时把控道路拥堵状况,针对道路拥堵情况提前做出预案和防范措施。减轻出行压力,提高居民幸福感。
With the rapid development of urbanization,traffic congestion has become an important problem that puzzles and hinders urban development. Most of the roads are partially congested and partially unobstructed. Accurate prediction of road congestion can better realize vehicle diversion and relieve traffic pressure. In this paper,we use VARMA(Vector Autoregressive Moving Average) and LSTM(Long-term and Short-term Memory Network) algorithms to model and analyze the congestion data of 57 roads near the Capital Airport. On this basis,the core idea of LSTM processing multiple time series is added to the multiple regression algorithm,so that the multiple regression algorithm has the ability to deal with multiple time series. Then the prediction accuracy and modeling complexity of the three algorithms are compared to find out the suitable algorithm for different scenarios. It is concluded that VARMA model is suitable for short-term accurate prediction,RNN model is suitable for long-term large-scale fluctuation prediction,and the modified multiple regression model is suitable for medium-term and long-term fast prediction. The algorithm and conclusion in this paper can better help the public security and traffic police to control the road congestion situation in time,and make plans and preventive measures in advance for the road congestion situation. Reduce travel pressure and improve residents' well-being.
With the rapid development of urbanization,traffic congestion has become an important problem that puzzles and hinders urban development. Most of the roads are partially congested and partially unobstructed. Accurate prediction of road congestion can better realize vehicle diversion and relieve traffic pressure. In this paper,we use VARMA(Vector Autoregressive Moving Average) and LSTM(Long-term and Short-term Memory Network) algorithms to model and analyze the congestion data of 57 roads near the Capital Airport. On this basis,the core idea of LSTM processing multiple time series is added to the multiple regression algorithm,so that the multiple regression algorithm has the ability to deal with multiple time series. Then the prediction accuracy and modeling complexity of the three algorithms are compared to find out the suitable algorithm for different scenarios. It is concluded that VARMA model is suitable for short-term accurate prediction,RNN model is suitable for long-term large-scale fluctuation prediction,and the modified multiple regression model is suitable for medium-term and long-term fast prediction. The algorithm and conclusion in this paper can better help the public security and traffic police to control the road congestion situation in time,and make plans and preventive measures in advance for the road congestion situation. Reduce travel pressure and improve residents' well-being.
引文
[1]崔承颖.基于累积Logistic模型的城市交通拥堵概率估计研究[D].北京:北京交通大学,2015.
[2]陈岳明,萧德云.基于跳转模型的路网交通流预测[J].控制与决策,2009,24(8):1177-1180+1186.
[3]陈韫.基于LSTM深度网络的城市道路短时交通状态预测模型研究[D].福建工程学院,2018.
[2]陈岳明,萧德云.基于跳转模型的路网交通流预测[J].控制与决策,2009,24(8):1177-1180+1186.
[3]陈韫.基于LSTM深度网络的城市道路短时交通状态预测模型研究[D].福建工程学院,2018.