基于售检票数据的城市轨道交通乘客分类

既有基于交通调查的乘客分类存在样本有限及分类标准主观性强等不足,本文以 乘客真实出行记录为基础,从“消费行为”视角构建客观的乘客分类指标及方法.为满足大规模 数据集处理需求,采用SPSS Modeler 软件对全样本乘客进行聚类.选取北京轨道交通连续1 个 月自动售检票(AFC)数据进行实证分析,结果显示：将乘客分为5 类时,聚类效果最佳;通过连 续5 个工作日聚类结果对比,验证了分类结果具有良好的稳定性.结合乘客分类结果进一步对 北京市轨道交通低峰折扣票价策略下不同类型乘客的出发时间转移弹性进行测定.该研究提 高了乘客分类客观性,能够为交通政策制定及运营策略评价提供方法支持.

Passenger Classification for Urban Rail Transit by Mining Smart Card Data

Traditional passenger classification methods based on traffic survey have drawbacks on limited sample and subjective standard, this paper constructs a new method and indexes in perspective of "consumer behavior" by using automatic fare collection (AFC) data. In order to meet the computation requirements of large data set, the SPSS Modeler is used to cluster the passengers. In case study, one month's AFC data of Beijing rail transit is applied and results shows that it is the best to cluster passengers in five classes, and the stability is verified by comparison with the clustering results in five consecutive days. The departure time transferring elasticities of different passenger types under pre-peak discount pricing strategy of Beijing transit are also analyzed. This study improves the objectivity of passenger classification and provides method support for traffic policy formulation and operation strategy evaluation.