基于关联规则和主成分分析的高铁旅客购票行为特征研究

High-speed railway passenger ticketing behavior characteristics based on association rules and principal component analysis

购票行为作为反映旅客出行选择的最终表现形式,是高速铁路客流分析的主要依据和客运组织工作的重要基础。针对当前高铁旅客出行需求“多元化”“差异化”的新特点,系统与全面地研究旅客购票行为特征可在切实提高客运服务质量的同时实现铁路运输企业效益最大化。依托大数据分析技术,以高速铁路历史售票信息数据为样本,提出一种基于关联规则和主成分分析的旅客购票行为特征研究方法,构建高铁旅客购票行为关键特征体系。通过改进Apriori算法中的“剪枝”操作,降低频繁项集生成时的数据读写负载,证明对客票数据等大规模数据集的关联规则挖掘效率平均提高23.7%。对关联结果从线路类型、出行时段、出行OD及退票时间4个方面进行分析,结果表明旅客购票决策及出行偏好特征在对于旅客购票行为全过程的研究中较旅客个体属性及社会环境等以往重点分析特征更具影响力。采用主成分分析法计算各购票行为特征权重,根据权重阈值选取提前购票时间、始发终到城市间距离、座席选择及购票决策总时间等10项特征构建购票行为关键特征体系,并以此提出基于旅客画像系统、预售期内票价动态浮动及差异化客票退改签等营销策略。研究结果对高铁运营部门准确把握客流变化趋势、优化客票组织策略及保障客运供需动态平衡提供了决策参考。

As the ultimate manifestation of passenger travel choice behavior,ticketing behavior is the main foundation for passenger flow analysis and passenger transport organization of high-speed railway.In view of the new characteristics of“diversification”and“differentiation”of the current high-speed railway passenger travel demand,a systematic and comprehensive study of the passenger ticketing behavior characteristics can effectively improve the quality of passenger transport services while maximize the benefits of railway transport enterprises.By taking the historical ticketing data of high-speed railway as an example and relying on the big data analysis technology,a method was proposed for passenger ticketing behavior characteristics based on association rules and Principal Component Analysis(PCA),and a key characteristic system of high-speed railway passenger ticketing behavior was constructed.The“pruning”operation in Apriori algorithm was improved,which reduces the data input and output load during the generation of frequent itemset and lifts the mining efficiency of association rules of large-scale datasets such as passenger ticket data by 23.7%on average.By analyzing the association rules from four aspects,including railway line,travel time,travel origin-destination(OD)and refund time,the results show that the characteristics of passenger’s ticketing decision and travel preference are more influential in the study of the whole process of passenger ticketing behavior than the previous key analysis characteristics such as passenger’s individual attribute and social environment.Each weight of ticketing behavior was calculated with PCA,and a key characteristic system of ticketing behavior was constructed including 10 characteristics,such as advance ticketing time,OD distance,seat type and decision time of ticketing which were selected according to the weight threshold.The high-speed railway passenger transport marketing strategies were put forward based on the system,including the precise marketing based on passenger portrait system,dynamic floating ticket price during pre-sale period and differentiated ticket refund and change strategy.The research results provide a decision-making reference for helping high-speed railway operation department grasp the trend of passenger flow accurately,optimize the ticketing strategy,and ensure the dynamic balance of passenger transport both supply and demand.