In the process of rail transit operation management, the characteristics and rules of time distribution of passenger flow between different stations directly determine the passenger flow organization scheme. The clarification of inter-station passenger flow time distribution types is beneficial to the reasonable configuration of passenger flow organization scheme. In this paper, we firstly analyze the classification characteristics of inter-station passenger flow from three perspectives of time, space and structure, and compress the search space of classification by spectral clustering method to achieve more accurate type classification. The classification results of different methods are compared using Silhouette Coefficient and Davies-Bouldin Index, and it is demonstrated that the proposed spectral clustering method has better classification results compared with other methods such as k-means. Taking Suzhou Metro 2020 data as an example, seven types of inter-station passenger flow time distribution are found by the proposed method, and the results can be applied to areas such as prediction model training.