基于多时间粒度的地铁出行规律相似性度量

Similarity Measurement of Metro Travel Rules Based on Multi-time Granularities

乘客出行规律对城市轨道交通运营管理至关重要,而不同时间粒度下观测到的客流规律差异较大。以往研究缺乏多时间粒度车站层级客流规律的量化研究。本文基于刷卡数据分析不同时间尺度下地铁出行规律的相似性。构建客流时间序列模型和相似性度量方法,使用连续五周北京地铁刷卡数据分别度量1 min到720min共16个时间粒度下,进站客流和OD客流与历史同期的相似性大小,并基于度量结果给出一定精度要求下预测短时进站量和OD量时的最小时间粒度推荐值,以综合相似性指标对全网车站的可预测等级进行划分。多角度统计分析结果表明,工作日客流与历史同期相似性较大,高峰比平峰、早高峰比晚高峰相似性大。

The regularity of passenger travel is important to the operation and management of urban rail transit.There is a great difference of observed passenger flow regularity in different time granularities.Given insufficient quantitative research in the past on the station-level passenger flow regularity at multi time scale,this paper analyzed the similarity of passenger travel rules based on smart card data at different time granularities.Firstly,the time series model and the similarity measure method were constructed.Then,the smart card data for five consecutive weeks of Beijing Metro were used to measure the similarity of inbound passenger flow and OD passenger flow during the same period of history over a scale of 16 time granularities from 1 minute to 720 minutes.Based on the results of similarity measurement,the minimum time granularity was recommended for the prediction of short term inbound traffic flow and OD passenger flow under certain precision requirements.Finally,the predictable levels of the stations of the whole network were classified by the comprehensive similarity index.The results of multi-perspective statistical analysis show a great similarity between passenger flow and historical flow during weekdays.The peak hour passenger flow similarity is greater than that of off-peak hours.The morning peak hour passenger flow similarity is greater than that of afternoon peak hours.