High-speed rail station entering and exiting passenger flow prediction method based on multi-task learning and large language model

By combining the ST-Attn-LLM framework with the spatiotemporal attention mechanism and the large language model, the problems of insufficient collaborative modeling and spatiotemporal feature extraction in high-speed rail inbound and outbound passenger flow prediction are solved, achieving high-precision passenger flow prediction and adapting to changes in complex traffic environments.

CN120806271APending Publication Date: 2025-10-17BEIJING JIAOTONG UNIV
0 Cites 0 Cited by

Patent Information

Application Number
CN202511011672.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-07-22
Publication Date
2025-10-17

AI Technical Summary

Technical Problem

When faced with complex spatiotemporal characteristics, the existing high-speed rail inbound and outbound passenger flow prediction models have problems such as a single prediction target, lack of collaborative modeling, insufficient spatiotemporal feature extraction capabilities, and lack of the ability to express cross-spatiotemporal causal mechanisms, making it difficult to accurately predict the dynamic changes of high-density passenger flow.

Method used

The ST-Attn-LLM framework based on multi-task learning and a large language model is adopted to extract the spatial dependencies and temporal dynamic features of inbound and outbound passenger flows through the spatiotemporal attention mechanism, and the large language model is used to enhance the features, thereby realizing information interaction between inbound and outbound flows and modeling causal influences across stations and time.

Benefits of technology

It significantly improves the accuracy and robustness of high-speed rail inbound and outbound passenger flow predictions, can effectively capture passenger flow change trends in complex traffic environments, and support intelligent scheduling and risk warning systems.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention discloses a multi-task learning and large language model-based high-speed rail in-and-out passenger flow prediction method. The method comprises the following steps: extracting in-and-out station flow information based on in-and-out station passenger flow data of a station in a historical time period; and on the basis of the inbound and outbound traffic information, predicting inbound traffic and outbound traffic by using a trained multi-task prediction model. Wherein the multi-task prediction model comprises a spatio-temporal feature extraction module, a spatio-temporal feature enhancement module and a prediction module, and the spatio-temporal feature extraction module is obtained based on a spatio-temporal attention mechanism to obtain time features and spatial features; a spatial-temporal feature enhancement module fuses the temporal features and the spatial features and obtains enhanced features by using a pre-trained large model, and the enhanced features are mapped into dense feature vectors; and the prediction module takes the dense feature vector as input to realize prediction of the inbound flow and outbound flow at the future moment. According to the invention, the accuracy and robustness of passenger flow prediction are obviously improved.
Need to check novelty before this filing date? Find Prior Art