论文部分内容阅读
短时交通流量预测,是交通系统信息化和智能化交通运输管理技术领域研究的关键问题.目前的方法对历史数据具有较高的依赖程度,或者具有较高的计算成本,或者不能有效反映实际中较复杂的交通网络及各结点之间的相互关系、以及依赖的不确定性,或者多种模型的组合使得预测方法较复杂.贝叶斯网是一种重要的概率图模型,本文以交通网络结构为基础,利用概率图模型在不确定性知识表示和推理方面的良好性质,考虑路口交通流量及其预测的时序依赖特征,构建了带有时序条件依赖关系的交通贝叶斯网.进而针对短时交通流量预测的实时性和高效性要求,提出了基于Gibbs采样的交通贝叶斯网近似概率推理算法,并进行交通流量的短时预测.实验结果表明,本文提出的交通贝叶斯网构建、近似推理以及相应的短时交通流量的预测方法,具有高效性、准确性和可用性.
Short-term traffic flow forecasting is a key issue in the field of transportation system informatization and intelligent transportation management technology.The current method has a high dependence on historical data, or has a high computational cost, or can not effectively reflect the actual The more complex traffic network and the interdependence between nodes, and the dependence on uncertainty, or a combination of multiple models make the prediction method more complex.Bayes network is an important probability graph model, this article Based on the traffic network structure, the traffic Bayesian network with time-dependent dependencies is constructed by taking advantage of the good properties of the probability map model in terms of uncertainty knowledge representation and reasoning. Considering the traffic flow at intersections and the timing dependency of its prediction, In order to meet the requirements of real-time and high efficiency of short-term traffic flow prediction, this paper proposes an approximate probability reasoning algorithm of traffic Bayesian network based on Gibbs sampling and short-term traffic flow prediction.The experimental results show that the proposed traffic Baye Network construction, approximate reasoning and the corresponding short-term traffic flow forecasting method, with high efficiency, accuracy and availability.