摘要
首先结合车头时距不稳定性的概念,对串车现象给出判定依据;分析串车问题产生的6种典型过程,建立串车问题数学模型,对串车发生的原因进行理论分析.然后,以自贡市38路公交线路实际数据为例,采用方差分析与回归分析筛选出7个串车形成影响因素,并对其重要性进行排序.最后,建立基于径向基函数神经网络的串车问题预测模型,对未来站的车头时距进行预测,并选取自贡市38路公交车的实时数据进行模型验证.在38 000余组数据中随机选择300组数据进行训练,对比30组测试数据.结果表明,学习得到的预测值与实际值偏差10%以内的样本点占90%,结果良好,证明了所建立的模型具有较好的适用性.
First,based on the concept of vehicle headway instability,the criterion for bus bunching phenomenon is proposed.By analyzing the six typical processes of the bus bunching problem,a mathematical model is established,with which the theoretical factors of bus bunching are analyzed.Then,taking the actual data of No.38 bus route in Zigong city for example,seven influencing factors are selected by using variance analysis and regression analysis,and are sorted by the importance.Finally,a forecasting model of the bus bunching problem based on radial basis function(RBF) neural network is established to forecast the headway of the future station.The real-time data of No.38 bus route in Zigong city are employed to validate the proposed model.The 300 groups of data randomly selected from more than 38 000 groups of real data are used for training.By comparing30 sets of test data,the results show that the samples with the errors between predicted and actual values within 10% account for 90%,which illustrate the result is good.It is proved that established model has good applicability.
出处
《东南大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2017年第6期1269-1273,共5页
Journal of Southeast University：Natural Science Edition
基金
国家重点研发计划资助项目(2016YFB0100906)
国家自然科学基金资助项目(61620106002
51308115)
关键词
公共交通
串车
排序
径向基函数神经网络
public transit
bus bunching
sort
radial basis function(RBF) neural network
