Identifying Subway Passenger Flow under Large-Scale Events Using Symbolic Aggregate Approximation Algorithm

Hainan Huang, Rongjie Zhang, Chengguang Xie, Xiaofeng Li

Research output: Chapter in Book/Report/Conference proceedingChapter

1 Scopus citations


Various social events, such as holidays, important sporting events, and major celebrations, may result in sudden large-scale passenger flows in certain sections and stations of urban rail transit systems. The sudden inbound passenger flows caused by these events can easily lead to continuous congestion of the subway network, which has a profound impact on the safety, reliability, and stability of a subway system. Because of the large magnitude of swipe data and the high dimensionality of time series, it is difficult to identify the emergence of such large passenger flows. Additionally, the recognition accuracy of the existing identification methods cannot meet the operational monitoring requirements. To address the above-mentioned issues, this paper proposes an optimized symbolic aggregate approximation (SAX) algorithm to identify historical sudden passenger flows caused by large-scale events around subways. Specifically, pre-set cluster types and dynamic time warping (DTW) are proposed to enhance the matching rate. Compared with the K-means method, the proposed method exhibits an average increase of 30% in mining accuracy, and the calculation time is shortened to one-sixteenth of the original value.

Original languageEnglish (US)
Title of host publicationTransportation Research Record
PublisherSAGE Publications Ltd
Number of pages11
StatePublished - Feb 2022


  • Big data
  • Data and data science
  • Public transportation
  • Rail transit systems
  • Subway
  • Urban transportation data and information systems

ASJC Scopus subject areas

  • Civil and Structural Engineering
  • Mechanical Engineering


Dive into the research topics of 'Identifying Subway Passenger Flow under Large-Scale Events Using Symbolic Aggregate Approximation Algorithm'. Together they form a unique fingerprint.

Cite this