| 基于换乘链断裂点时空信息的公交换乘行为识别 |
| |
| 引用本文： | 靳海涛,金凤君,陈卓,王姣娥,杨宇.基于换乘链断裂点时空信息的公交换乘行为识别[J].交通运输工程学报,2018,18(5):176-184. |
| |
| 作者姓名： | 靳海涛 金凤君 陈卓 王姣娥 杨宇 |
| |
| 作者单位： | 1.中国科学院地理科学与资源研究所区域可持续发展分析与模拟重点实验室,北京 1001012.北京市交通信息中心,北京 1001613.中国科学院大学资源与环境学院, 北京 100049 |
| |
| 基金项目： | 国家自然科学基金项目41771134中国科学院战略性先导科技专项项目XDA19040403 |
| |
| 摘 要： | 分析了公交智能卡数据挖掘中转乘行为、通勤出行、非通勤出行识别的改进方法, 将关注点从公交换乘过程信息转移到换乘链间隔期的持续时长和空间位移信息, 以换乘链断裂时长和位移2个维度计算公交换乘链断裂点概率, 制作工作日和非工作日断裂点时空变量联合概率分布矩阵, 对比了这2种分布的差异;检验了断裂时长序列和断裂位移序列的稳定性, 标识了2条曲线的突变点和拐点, 用于推断转乘引起的转移距离和转乘时长的阈值参数;对工作日和非工作日差值时长序列曲线进行移动平均滤波处理, 使得曲线的突变与极值之间的关联能够解释转乘、通勤出行和非通勤出行3种行为与通勤和非通勤出行之间的关联;采用北京市整个一周的地面公交和地铁系统样本数据对方法进行验证, 并根据时间序列和位移序列曲线确定样本数据中常见公交换乘行为的阈值参数。分析结果表明:断裂点时空信息对样本数据中的换乘行为能提供更合理的识别分类参数;持卡人在站点间转移的容忍距离约为1.6km;断裂点转乘时长与非通勤出行的断裂时长临界点为22~48min;非通勤出行和通勤出行的时长临界点约为478min, 非通勤出行断裂点最大概率时长为140min;通勤出行的断裂时长接近期望值为601且标准差为44的正态分布;基于新方法得出的参数改善了公交出行活动的识别率, 转乘行为、通勤出行与非通勤出行的识别率分别提高了16.1%、4.2%、6.2%。可见, 换乘链断裂点的时间信息和空间信息不但可作为公交换乘行为识别的依据, 还可能带来更好的识别效果。
|
| 关 键 词： | 城市交通 公交换乘行为 换乘链断裂点 时空信息 智能卡数据 数据挖掘 |
| 收稿时间： | 2018-03-12 |
Commute activity identification based on spatial and temporal information of transit chaining breaks |
| |
| JIN Hai-tao,JIN Feng-jun,CHEN Zhuo,WANG Jiao-e,YANG Yu.Commute activity identification based on spatial and temporal information of transit chaining breaks[J].Journal of Traffic and Transportation Engineering,2018,18(5):176-184. |
| |
| Authors: | JIN Hai-tao JIN Feng-jun CHEN Zhuo WANG Jiao-e YANG Yu |
| |
| Affiliation: | 1.Key Laboratory of Regional Sustainable Development Modeling,Institute of Geographic Sciences and Natural Resources Research,Chinese Academy of Sciences, Beijing 100101, China2.Beijing Transportation Information Center, Beijing 100161, China3.College of Resources and Environment,University of Chinese Academy of Sciences, Beijing 100049, China |
| |
| Abstract: | An approach to improve the recognition of transfers, working commutes, and nonworking commutes in smartcard data mining was introduced.The study focus was shifted from the information of transit processes to the durations and displacements between the transit chaining breaks.The probabilities of transit chaining breaks were calculated by two dimensions of the break durations and displacements, and a joint probability distribution matrix of spatial andtemporal variables for workdays and non-working days was made.The differences between the two types of distribution were compared.The stabilities of the break duration sequences and break displacement sequences were examined.The mutation points and turning points of the two curves were marked to infer the important threshold parameters for the transferring durations and displacements generated by the transfers.A moving average filter was utilized to smooth both workdays and non-working days curves of margin duration values.The relationship between the mutation and extremes of the curve was explained for the three types of commute activities relating to the transfers, working commutes, and non-working commutes.The approach was verified by a weeklong sample dataset of the Beijing bus and subway system.The threshold parameters of the common commute activities in the dataset were determined according to the time series and the displacement sequence curve.Analysis result shows that the spatial and temporal information at the breaks can provide more reasonable identification parameters for the commute activities.A tolerance distance of approximately 1.6 km between the transit connections is found among the cardholders.The threshold of transit break duration between the transferring and non-working commutes is 22-48 min.The threshold of working and non-working commutes is approximately 478 min, and the maximum probability of non-working duration is 140 min.The transit chaining break durations of working commutes fall into a normal distribution with an expected value of 601 and a standard deviation of 44.The parameters generated by the new approach lead to an improvement in commute activity recognition, the recognition rates of the transfers, working commutes and non-working commutes increse by 16.1%, 4.2%and 6.2%, respectively.So the spatial and temporal information of transit chaining breaks can not only provide the basis for the commute activity identification, but also achieve better recognition results. |
| |
| Keywords: | |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《交通运输工程学报》浏览原始摘要信息 |
|
点击此处可从《交通运输工程学报》下载全文 |
|
