Abstract: A refined classification method of transit riders under bus one-ticket system was proposed. Firstly, due to the characteristics of the bus one-ticket system, only boarding information is available. The characteristic indexes of the riders, namely, the travel intensity, temporal indicators, spatial indicators and peak-hour indicators, were selected as the indexes of the classification. The correlation analysis was employed to screen indicators. Then, Two-Step clustering was used to cluster transit riders into clear clusters and unclear clusters. Moreover, feature indicators were selected again for clear clusters and unclear clusters for secondary clustering to achieve the refined classification of passengers. A case study on the transit riders data of the one-ticket bus system in Guangzhou shows that the proposed method can effectively classify passengers, and the results of each class depict the travel characteristics of passengers in detail. The proposed classification method is effective and stable.

Key words: traffic engineering; one-ticket bus; transit rider classification; travel characteristics; combinatorial clustering

摘要： 构建了一种可应用于公交一票制下的乘客精细化分类方法。针对公交一票制仅有上车信息的特点，从乘车强度、时间特性、空间特性及高峰特性构建分类指标，并利用相关性分析进行指标筛选；然后应用Two-Step聚类对乘客进行聚类，将乘客分为清晰簇和非清晰簇，并对清晰簇和非清晰簇再次选取特征指标进行2次聚类以实现乘客的精细化分类。根据广州市的算例表明，该方法可以对乘客进行有效分类，每类结果对乘客的出行特征进行了细致地刻画，分类方法有效且稳定。

关键词: 交通工程；公交一票制；乘客分类；出行特征；组合聚类