Abstract:In order to explore the passenger flow pattern of urban rail transit stations, AFC data was used to construct passenger flow indicators, and a station passenger flow recognition model based on K-means clustering algorithm was proposed. Taking the AFC data of Chongqing Rail Line 3 for 1 month as an example, the clustering results of different passenger flow indexes and comprehensive multi-variable indexes during working days, weekends and holidays are discussed. The results show that: passenger flow indexes in different periods can promote station passenger flow identification; When the station passenger flow pattern is divided into 7 categories, the clustering efficiency is the best; by comparing the clustering results for 1 week and 1 month, the classification results are verified to have good stability. The characteristics of the station passenger flow are summarized and combined with the characteristics of the resulting data and the actual situation of the station.