W-shape作为站点客流序列聚类方法的设计与实现  

W-shape:Design and Implementation of Site Passenger Flow Sequence Clustering Method

作　　者：张一帆 吕永波[1] 马继辉 吕万钧 ZHANG Yifan;LV Yongbo;MA Jihui;LV Wanjun(Institute of Systems Engineering and Control,Bijing Jiaotong University,Beijng 100044,China)

机构地区：[1]北京交通大学系统工程与控制研究所,北京100044

出　　处：《综合运输》2020年第5期75-82,共8页China Transportation Review

基　　金：国家自然科学基金(61872036);国家重点研发计划项目资助(2017YFC0804900)。

摘　　要：公交站点客流在一天中的分布可看做一种时间序列,但不同于一般的时间序列,公交客流序列往往有着较为明显的早、晚高峰。对站点客流序列进行有效的识别与分类能够对公交排班及客流预测起到支持作用。本文提出了W-Shape,一种新的客流聚类方式。可迭代可扩展且能够有效根据客流分布划分类别。针对公交客流时间序列的一般性质,本文设计了加入正则项的相关性衡量方式weightSBD,这是一种有效地衡量时间序列间距离的方式,并与其他时间序列聚类方式进行对比。比较结果表明,W-shape是一种有效的聚类模型,可用于公交客流序列的识别与聚类中。The distribution of bus station’s passenger flow can be seen as a time series, but unlike the general time series, the bus passenger flow sequence often has obvious early and late peaks. Effective identification and classification of the site passenger flow sequence can support bus scheduling and passenger flow forecasting. This paper proposes W-Shape, a new method of passenger flow sequences clustering, which is iteratively scalable and able to effectively classify categories based on passenger flow distribution. Aiming at the general nature of bus passenger flow time series, a weight-SBD method for measuring the correlation of regular items was designed. This is an effectively method to measure the distance between time series compared with other time series clustering methods. The results of experiment show that W-shape is an effective clustering model and can be used in the identification and clustering of bus passenger flow sequences.

关 键 词：城市交通 客流序列聚类 数据挖掘 W-shape weight-SBD 

分 类 号：U491.17[交通运输工程—交通运输规划与管理]