Chand Sai
Research Centre for Integrated Transport Innovation (rCITI), School of Civil and Environmental Engineering, University of New South Wales, Sydney, NSW 2052, Australia.
Entropy (Basel). 2021 Feb 3;23(2):188. doi: 10.3390/e23020188.
Predictability is important in decision-making in many fields, including transport. The ill-predictability of time-varying processes poses severe problems for traffic and transport planners. The sources of ill-predictability in traffic phenomena could be due to uncertainty and incompleteness of data and models and/or due to the complexity of the processes itself. Traffic counts at intersections are typically consistent and repetitive on the one hand and yet can be less predictable on the other hand, in which on any given time, unusual circumstances such as crashes and adverse weather can dramatically change the traffic condition. Understanding the various causes of high/low predictability in traffic counts is essential for better predictions and the choice of prediction methods. Here, we utilise the Hurst exponent metric from the fractal theory to quantify fluctuations and evaluate the predictability of intersection approach volumes. Data collected from 37 intersections in Sydney, Australia for one year are used. Further, we develop a random-effects linear regression model to quantify the effect of factors such as the day of the week, special event days, public holidays, rainfall, temperature, bus stops, and parking lanes on the predictability of traffic counts. We find that the theoretical predictability of traffic counts at signalised intersections is upwards of 0.80 (i.e., 80%) for most of the days, and the predictability is strongly associated with the day of the week. Public holidays, special event days, and weekends are better predictable than typical weekdays. Rainfall decreases predictability, and intersections with more parking spaces are highly predictable.
可预测性在包括交通运输在内的许多领域的决策中都很重要。时变过程的不可预测性给交通和运输规划者带来了严重问题。交通现象中不可预测性的来源可能是由于数据和模型的不确定性和不完整性，和/或由于过程本身的复杂性。一方面，交叉路口的交通流量计数通常是一致且重复的，但另一方面，其可预测性可能较低，因为在任何给定时间，诸如撞车和恶劣天气等异常情况都可能极大地改变交通状况。了解交通流量计数中高/低可预测性的各种原因对于更好的预测和预测方法的选择至关重要。在此，我们利用分形理论中的赫斯特指数度量来量化波动并评估交叉路口进近交通量的可预测性。我们使用了从澳大利亚悉尼的37个交叉路口收集的一年数据。此外，我们开发了一个随机效应线性回归模型，以量化诸如星期几、特殊活动日、公共假日、降雨量、温度、公交站和停车道等因素对交通流量计数可预测性的影响。我们发现，大多数日子里，信号控制交叉路口交通流量计数的理论可预测性超过0.80（即80%），并且可预测性与星期几密切相关。公共假日、特殊活动日和周末比典型工作日更具可预测性。降雨会降低可预测性，而停车位较多的交叉路口具有较高的可预测性。
