A Congestion Index Prediction Method Combining Road Network Topology Structure and Semantic Association

Abstract

A congestion index prediction method combining road network topology and semantic association, comprising the following steps: (1) establishing an undirected graph based on the spatial topology of the road network; (2) first calculating the similarity between road historical congestion index data, Then build a weighted undirected graph based on the similarity, and finally embed the weighted undirected graph to obtain the semantic vector representing the road; (3) Extract short-term congestion index change features based on graph convolutional network, and extract long-term congestion based on cyclic neural network Exponential change features, based on which the road semantic vector is fused to establish a prediction model. The present invention simultaneously considers the spatial topology association and historical semantic association of the road network, and improves the predictive ability of the model; adopts the graph convolution network to model the road network topology structure, and adopts the graph embedding to model the road network semantic association, so that the road network topology Structural and semantic associations can be handled by deep neural networks.

Description

technical field [0001] The invention relates to machine learning technology, in particular to a congestion index prediction method. Background technique [0002] The intelligent transportation system can collect the average vehicle speed of the road through coils, microwaves, cameras and other equipment, and then calculate the road congestion index, and the congestion index prediction refers to predicting its future congestion index based on the historical congestion index of the road. Congestion index prediction is of great significance to travel planning and traffic control. [0003] Congestion index prediction methods mainly include knowledge-driven methods and data-driven methods. The knowledge-driven method is a more traditional method, which mainly realizes prediction by simulating the operation of vehicles. A data-driven approach is a method to achieve predictions based on advanced machine learning techniques. Since the congestion index is a kind of time series dat...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the shortcomings of the poor prediction performance of the existing congestion index prediction methods, the present invention proposes a congestion index prediction method that combines road network topology and semantic association with strong prediction performance

Images