智能系统学报2024，Vol.19Issue(6)：1518-1527,10.DOI:10.11992/tis.202309006
基于图卷积神经网络的最短路径距离估计方法
Road network shortest distance estimation method based on graph convolutional networks
孟祥福 1崔江燕 1邓敏超1
作者信息
- 1. 辽宁工程技术大学 电子与信息工程学院,辽宁 葫芦岛 125105
- 折叠
摘要
Abstract
Improving the accuracy of estimating the shortest path distance while reducing model training time is crucial.Existing methods for embedded shortest path distance estimation often take too long to train or sacrifice accuracy to save time.To solve these problems,an encoder-decoder framework has been designed to estimate the shortest distance in a road network by analyzing existing embedded systems-based shortest-path distance estimation methods.The core process is broken down into three parts:embedding method,sampling method,and model training.A road network ver-tex embedding method,RGCNdist2vec,leverages road graph convolutional networks to capture the structural informa-tion of the road network.For model training,a three-stage sampling method using graph logical partitioning is designed to select a small number of high-quality samples.Experiments conducted on four real road network data sets demon-strate that the proposed model achieves higher estimation accuracy while reducing training time by nearly four times compared to existing baseline models.关键词
最短路径距离计算/图神经网络/数据采样/表示学习/图卷积网络/图分区/深度学习/拓扑结构Key words
shortest path distance computation/graph neural networks/data sampling/representation learning/graph convolutional networks/graph partitioning/deep learning/topology分类
计算机与自动化引用本文复制引用
孟祥福,崔江燕,邓敏超..基于图卷积神经网络的最短路径距离估计方法[J].智能系统学报,2024,19(6):1518-1527,10.
