Abstract:

Smart card data provides the data basis for the study of long-term travel regularity of public transport passengers. Based on the smart card data from April to May 2018 in Beijing, this study constructed the passenger activity chain in three steps, including extracting the passenger activity location, inferring the residence location, and identifying the activity type. The PrefixSpan algorithm was used to extract the frequent sequence patterns of activity chains for regular, senior and student card users. Levenshtein distance was applied to measure the day-today similarity of the three types of passengers' daily activity chain. The results show that about 70% of users in each type have a symmetrical pattern of frequent activity sequences. The similarity of regular card and student card users is higher than senior card users, with an average of 0.645, 0.649, and 0.530, respectively. For all the three types of users, the differences in day-to-day activity chain sequences between workdays and weekends are larger, but the similarities within workdays or within weekends are higher. This study helps to quantitatively analyze the regularity of passenger travel and activity and provides evidence for scientific optimization of public transport services.

Key words: intelligent transportation, similarity, sequence mining, public transport passengers, PrefixSpan algorithm

摘要：

刷卡数据为研究公共交通乘客长期出行规律提供了数据基础.利用北京市2018年 4~5月的刷卡数据，通过提取乘客活动地，推断居住地位置和识别活动类型3个步骤构建乘客个体活动链；基于PrefixSpan算法提取普通卡、老年卡、学生卡乘客活动链的频繁序列模式，采 用Levenshtein距离度量3类乘客活动链日维度的相似性.结果表明：每类用户中约70%乘客的频繁活动序列是对称模式；普通卡和学生卡用户的相似性高于老年卡用户，平均值分别为 0.645、0.649和0.530；3类乘客的工作日与非工作日活动链具有明显差异，而工作日之间或非工作日之间具有较高相似性.本文有助于定量解析公共交通乘客的出行活动规律，为科学优化公共交通服务提供依据.

关键词: 智能交通, 相似性, 序列挖掘, 公共交通乘客, PrefixSpan算法