A spatiotemporal heterogeneous decoupling method based on multi-modal traffic flow prediction

By constructing decoupling gates, residual graph convolution modules, and temporal flow modules, multi-scale and spatiotemporal correlations are captured, solving the problem of ignoring multi-modal relationships in traditional traffic flow prediction methods and achieving higher prediction accuracy and applicability.

CN117558125BActive Publication Date: 2026-06-02HANGZHOU DIANZI UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
HANGZHOU DIANZI UNIV
Filing Date
2023-11-10
Publication Date
2026-06-02

AI Technical Summary

Technical Problem

Traditional traffic flow prediction methods typically focus on a single traffic mode, neglecting the potential relationships and spatiotemporal heterogeneity between various traffic modes in modern transportation systems, which limits the accuracy of predictions.

Method used

A spatiotemporal heterogeneous decoupling method based on multi-modal traffic flow prediction is adopted. By constructing decoupling gates, residual graph convolution modules, temporal flow modules, and jump connection layers, multi-scale and spatiotemporal correlations are captured, and complex relationships between multiple traffic modes are handled.

Benefits of technology

It effectively handles the complex relationships between various traffic modes, improves the accuracy and applicability of traffic flow forecasting, and can better reflect the diversity and complexity of urban transportation systems.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117558125B_ABST
    Figure CN117558125B_ABST
Patent Text Reader

Abstract

The application belongs to the technical field of traffic flow prediction, and discloses a space-time heterogeneous decoupling method based on multi-mode traffic flow prediction, which comprises the following steps: step one, traffic flow data preprocessing; step two, constructing a decoupling gate to decouple input signal flow into multi-scale components; step three, constructing a gate-controllable residual graph convolution module; step four, constructing a time flow module for capturing long-term and short-term time dependencies; step five, a jump connection layer and a regression prediction module for alleviating gradient disappearance. The application can effectively handle the complex relationship between various different traffic modes, making it more suitable for modern urban traffic systems. By considering the interaction between different modes, the prediction can better reflect the diversity and complexity of urban traffic, providing more comprehensive information for traffic decision-making. It can meet the needs of various traffic flow prediction tasks. It helps better understand and simulate the dynamic changes of urban traffic systems.
Need to check novelty before this filing date? Find Prior Art