Urban traffic network accident prediction method based on graph representation learning and feature fusion

Through the method of graph representation learning and feature fusion, combined with graph neural network and comparison learning, the problem of long-tail distribution data in urban traffic networks is solved, the prediction accuracy and generalization ability of the model are improved, and efficient urban traffic accident prediction and planning are achieved.

CN120355237AActive Publication Date: 2025-07-22SUZHOU UNIV
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
CN202510813232.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-06-18
Publication Date
2025-07-22
Estimated Expiration
2045-06-18

AI Technical Summary

Technical Problem

In the process of accident prediction of urban transportation networks, the existing technology has problems of insufficient long-tail distribution data, feature fusion and model generalization capabilities, resulting in low prediction accuracy and inefficiency.

Method used

The method of graph representation learning and feature fusion is adopted, and node features, edge features and intersection angle information is extracted, combined with graph neural network and comparison learning, model parameters are optimized, and message delivery mechanism and deep learning technology are used to predict.

Benefits of technology

It improves the accuracy of accident prediction and generalization ability of models, can better handle imbalanced data in complex urban transportation networks, and achieve efficient prediction and planning.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120355237A_ABST
    Figure CN120355237A_ABST
Patent Text Reader

Abstract

The invention discloses an urban traffic network accident prediction method based on graph representation learning and feature fusion, which comprises the following steps: (1) processing urban traffic network accident data, and extracting node features, edge features and road direction and intersection angle information; (2) performing fusion embedding on the node features, the edge features, the road direction features and the intersection angle features to generate graph structure features; (3) constructing a graph neural network model based on a message passing mechanism, learning graph structure features through a graph convolutional layer, and extracting deep features of the traffic network; (4) training accident data in long-tail distribution by adopting a comparative learning method, and optimizing model parameters; (5) predicting accidents in the urban traffic network by using the trained model, and outputting a classification result of accident danger degrees; the invention provides a more powerful tool and method for urban traffic accident prediction and urban traffic road planning.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Traffic flow prediction method based on dynamic graph neural network

    CN114120652A

  • Intelligent traffic system anomaly prediction method and device and storage medium

    CN116307033A

  • Road traffic safety prediction method based on dynamic graph attention space-time network

    CN117746628A

  • Traffic accident prediction method based on graph attention network

    CN118053095A

  • Network embedding method based on attention mechanism

    CN118094252A