基于公交IC卡数据的上车站点推算研究  

Boarding Stop Inference Based on Transit IC Card Data

作　　者：马晓磊[1] 刘从从[1] 刘剑锋[2] 陈锋[3] 于海洋[1] 

机构地区：[1]北京航空航天大学车路协同与安全控制北京市重点实验室,北京100091 [2]北京城建集团有限责任公司,北京100037 [3]北京交通发展研究中心,北京100073

出　　处：《交通运输系统工程与信息》2015年第4期78-84,共7页Journal of Transportation Systems Engineering and Information Technology

基　　金：国家自然科学基金(51408019;51308021);北京市科技新星计划项目(z151100000315048)

摘　　要：为了分析城市公交乘客的出行特征,本文利用公交IC卡及GPS数据对公交IC卡乘客上车站点推算进行研究.针对安装车载GPS设备的车辆,运用GPS数据与IC卡数据融合算法进行推算;对于无车载GPS设备的情况,为适应一票制IC卡数据挖掘,对贝叶斯决策树算法进行改进,允许节点跳跃,推算上车站点,并且利用Markov链特性降低算法的运算复杂度.同时,本文以北京公交数据为例,对提出的两种方法进行验证.结果表明,利用本文提出的方法推算上车站点,3站之内误差的准确率达到90%以上,算法在兼顾算法精度的同时合理地控制了运算复杂度,可以实际运用于城市公交系统.In order to analyze urban bus passengers＇ travel characteristic, this paper proposes several data mining algorithms for boarding stop inference based on IC card and GPS data. For those buses with GPS devices, a data-fusion method with GPS data is developed to estimate individual passenger＇ s boarding stop. For those buses without GPS devices, an improved Bayesian decision tree algorithm with varying steps is presented to calculate the likelihood of each possible boarding stop. In addition, Markov Chain optimization technique is applied to reduce the computational complexity. Empirical data from Beijing transit route are used to validate the effectiveness of the proposed algorithms. The results demonstrate that the accuracy of identified boarding stop can be guaranteed and the algorithm complexity can be well controlled to meet the requirements of practical application. As a result, the methods can be widely adopted for urban public transportation system.

关 键 词：城市交通 公共交通 上车站点推算 贝叶斯决策树算法 IC卡数据 GPS数据 

分 类 号：U491[交通运输工程—交通运输规划与管理]