Journal Article

A method for short-term passenger flow prediction in urban rail transit based on deep learning

Li, Tiezhu, Tu, Ran, Bo, Yiyong, Liu, Hui, Dong, Ningning, Lin, Fei, Liu, Tianhao

2024

Overview

Short-term passenger flow prediction is a critical component of urban rail transit operations. However, predictions of passenger flow are mostly focused on one station, and land use, which has a substantial impact on passenger flow variation, has not been taken into account. A model termed the temporal-spatial network long short-term memory model (TNS-LSTM) is developed to solve the forecasting gap for the metro inbound/outbound passenger flow. The model introduces the spatial characteristics of the land use by extracting the point of interest (POI) data instead of merely considering temporal characteristics and network characteristics. The spatial-temporal network matrix is designed through the K-Means clustering model, extraction for temporal characteristics analysis for land use, and establishment of an origin-destination station matrix. Furthermore, the prediction of short-term passenger flow is implemented for multiple stations in the metro network. Finally, a case study based on actual data from the Nanjing metro is carried out, and the results demonstrate that the proposed model can not only avoid the complexity of constructing the numerous models for each station in urban rail transit but also improve the prediction accuracy and save a substantial amount of time.

• Publisher: Springer US,Springer Nature B.V
• Subject: 1229: Multimedia Data Analysis for Smart City Environment Safety / Cluster analysis / Clustering / Computer Communication Networks / Computer Science / Critical components / Data Structures and Information Theory / Deep learning / Land use / Multimedia Information Systems / Passengers / Special Purpose and Application-Based Systems / Subway stations / Urban rail / Vector quantization