基于粒子群小波神经网络的公交到站时间预测

Prediction Model of Bus Arrival Time Based on Particle Swarm Optimization and Wavelet Neural Network

公交到站时间的实时预测是公交出行信息发布、公交出行诱导、公交动态调度的关键技术.基于公交车辆运行特性分析,将公交到站时间分为路段运行时间和站点停靠时间两部分,并考虑工作日与周末的运行特性差异,最后结合迭代思想提出利用粒子群小波神经网络模型预测公交到站时间.实例分析表明:粒子群算法能有效降低小波神经网络模型的训练误差;结合迭代法使用公交车上一站运行时间作为预测输入能够有效提高预测精度;该预测模型对于公交车在工作日和周末到站时间的预测均能达到较高的精度,平均绝对百分比误差分别为10.82%和9.85%.

Real-time bus arrival time prediction is the key technology of bus travel information release, bus trip guidance, and bus dynamic scheduling. Based on the characteristic analysis of bus operation, the bus arrival time is divided into section running time and platform docking time. With the consideration of differences between running characteristics of the weekday and weekend, a forecasting model is proposed based on iterative thinking, particle swarm optimization and wavelet neural network to forecast bus arrival time. Example analysis shows that the particle swarm optimization can effectively reduce the training error of wavelet neural network model. Combined with the iterative method, the use of bus running time as forecast input can effectively improve the prediction accuracy. The bus arrival time prediction model is built in this paper can reach high precision on weekday and weekend, and mean absolute error is 10.82% and 9.85%.