基于贝叶斯组合模型的短期交通量预测研究

Study of short-term freeway traffic flow prediction based on BAYESIAN combined model

提出一种新的贝叶斯组合神经网络模型并将其应用于短期交通流量的预测。模型通过实时跟踪模型的预测表现,根据研究提出的分配算法不断调整模型的信用值,从而挑选并组合得到精度更高的预测模型。介绍了该模型的基本原理及在示范路网中的实际应用,通过选取反向传播神经网络和径向基函数神经网络,用以构造贝叶斯组合模型,并在测试数据集中进行了性能比较。计算结果表明:模型的预测性能整体上优于单一的神经网络模型,并且确保了模型预测的稳定性。

Method named BAYESIAN combined neural network model is proposed for short term traffic flow prediction in this paper. Such a model tracks the prediction performance of the combined models, and adjusts their credit values according to a proposed set of assignment algorithm so as to select the models with higher accuracy for combination. The basis theory of this model and its application in a test network are expounded. Back propagation (BP) neural network and radial basis function (RBF) neural network are selected to construct the combined model, which are applied to a test data set for performance comparison. It is found that the performances of combined model are better than those of the singular neural network model. More important, it is the characteristic of the combined model to track prediction model performance online and to always combine the better performing predictors for a stable prediction.