OBD支持下公交车到达时间的回归预测方法

Regression prediction method of bus arrival time supported by OBD

公交车到达时间的精准预测是共享公交专用道研究的基础,为此基于车辆车载自动诊断系统(OBD)数据提出一种基于证据理论优化高斯过程回归(DS-GPR)的公交车到达时间预测方法。首先分析影响公交车到达时间的影响因素,随后通过历史融合数据训练得到模型的超参数,并且最终通过D-S证据理论为高斯过程的回归输出分布分配全局信任度,得到当前时刻信任度最优的输出值。最后通过实例对DS-GPR模型的预测性能进行对比分析,证明了DS-GPR模型对公交车到达时间的良好预测性能。

The accurate prediction of bus arrival time is the basis of the study of shared bus lanes.Based on the vehicle on board diagnostics(OBD)data,Gaussian process regression(DS-GPR)for bus arrival time prediction with evidence theory optimization is proposed.Firstly,the influencing factors of bus arrival time are analyzed,and then the hyper parameters of the model are obtained by historical data training.Then,the optimal output values are selected from the regression output distribution of Caussian process with the global trust degree allocated according to D-S theory.Finally,the prediction performance of DS-GPR model is analyzed by comparing with other models.It is proved that DS-GPR model has good prediction performance for bus arrival time.