基于公交浮动车数据的城市主干道交通拥堵预测

交通拥堵预测是解决交通拥堵问题的前提。针对速度特性分析不全面的问题，基于公交浮动车数据，在速度时间相关性和空间相关性分析的基础上，加入了公交流量和时间占有率两个特征，提出了考虑时空特性和公交车流特性的改进粒子群优化的径向基函数神经网络(particle swarm optimization-radial basis function, PSO-RBF)速度预测模型。通过比较预测结果与速度阈值，得到城市主干道的交通拥堵情况。结果表明，与只考虑时空特性的预测结果相比，所提出的基于时空特性和公交车流特性的预测方法，可使模型预测的均方根误差(root mean square error, RMSE)、平均绝对误差(mean absolute error, MAE)分别降低13.58%、12.63%,决定系数达92.39%。同时，实例验证了改进的PSO-RBF神经网络模型的预测精度要优于标准的PSO-RBF神经网络。

Traffic Congestion Prediction of Urban Trunk Roads Based on Floating Bus Data

Traffic congestion prediction is the prerequisite to solve the problem of traffic congestion. In order to solve the problem of incomplete analysis of speed characteristics, the spatio-temporal characteristic of speed was analyzed based on floating bus data firstly, then the two characteristics of bus traffic and time occupancy were used to propose the improved particle swarm optimization radial basis neural network (PSO-RBF) speed prediction model. Finally, through comparative prediction results and speed thresholds, traffic congestion of the urban main roads was obtained. The results show that compared with the prediction results that only consider the spatio-temporal characteristic, the root mean square error (RMSE) and mean absolute error (MAE) of the model are reduced by 13.58% and 12.63%, respectively, and the determination coefficient reaches 92.39%. At the same time, the example verifies that the prediction accuracy of the improved PSO-RBF neural network is better than that of the standard PSO-RBF neural network.