面向海量公交刷卡数据的站点客流分析方法

Passenger Flow Analysis of Bus Stations on Massive Bus Card Data

公交行业的发展,产生了海量多元的公交IC卡刷卡数据,为行业应用提供快速、准确的站点客流量统计一直是智能公交建设的重点。以往对客流量的研究只是进行了简单的数据统计,准确度不高,提出的并行算法在海量数据规模下不具备水平扩展能力。针对此问题,论文通过分析海量多元数据的特征,面向公交刷卡数据提出了一种刷卡时间的聚类方法,不仅可在分钟级完成一周数据的计算,并根据换乘的时间差和距离规则约束提高了计算的准确性。论文工作在Hadoop MapReduce上进行了实现,分时客流量的计算方面随数据规模增大具有可扩展性,单位数据规模的计算执行时间保持相对稳定,并且分析结果具有较高的准确性。

In the public traffic business,massive and diverse bus IC card data have been generated,and it is the key point of the intelligent transport to provide quick and accurate passengers flow analysis of bus stations.In the past the study of traffic is just a simple data statistics and the accuracy is not high,as well as the parallel algorithm under the massive data size does not have the ability of horizontal extension.Aiming at this problem,through the analysis of the characteristics of huge amounts of multivariate data,a kind of clustering algorithm is developed based on the massive bus card data,the calculation of a week of data can be completed in minutes,and the calculation accuracy is improved according to the rules of the time difference and distance constraints.The computing of time-sharing traffic has a good expansibility basis on Hadoop MapReduce when the data scale is increasing,the execution time remains relatively stable in the unit data scale,and the results of the analysis has high accuracy.