城市道路交通拥堵预测及持续时间研究

Research on the Urban Traffic Congestion Predication and Time of Duration

针对城市道路交通拥堵及持续时间辨识问题,提取平均旅行速度、平均旅行时间、平均延迟时间、早晚高峰、星期数等交通拥堵关键影响因素,构建了基于MapReduce的多元对数线性回归交通拥堵预测模型和基于生存分析的交通拥堵持续时间模型,并利用上海快速路段交通数据集进行模型有效性验证。试验结果表明,拥堵预测模型预测值与实际值拟合度在0.96以上,能较好地量化道路交通运行拥堵程度;拥堵持续时间模型可以辨识出拥堵分布和持续时间特征,为制定交通拥堵的控制和疏导策略提供指导性建议。

In order to accurately predict the real-time status of urban traffic congestion and time of duration, in this paper the key factors of traffic congestion are extracted such as travel speed, travel peak time, delay time, morning peak, evening peak and weekdays, and the traffic congestion prediction model is established based on MapReduce multivariate logarithm linear regression and traffic congestion duration model through survival analysis method. The experiments are verified by traffic big data of Shanghai expressway sections. The results show that the fitness of congestion prediction model between actual and estimated values is above 0.96, which can better quantify the degree of traffic congestion; besides, the congestion duration model can identify the characteristics of traffic congestion distribution and duration, which will provide guidance for traffic control and strategies.