Traffic Forecasting using Temporal Line Graph Convolutional Network: Case Study

Abstract

Traffic forecasting is imperative to Intelligent Transportation Systems (ITS), and it has always been considered as a challenging research topic, due to the complex topological structure of the urban road network and the temporal stochastic nature of dynamic change. Popular sports events attract vast numbers of spectators travelling to the event, which will have a substantial effect on ITS, showing peaks on the network that can collapse a smart city's ITS. In this paper, we tackle traffic forecasting and use the Doha network in Qatar and the FIFA World Cup 2022 (FWC 2022) event as a case study. We propose a novel technique for embedding road network graphs into a Temporal-Graph Convolutional Network. The embedding process includes a modification to the graph weights based on graph theory and the properties of the line graph. Extensive simulations are carried out on a real-world calibrated dataset from Doha's road network. Our Temporal Line Graph Convolutional Network (TLGCN) proposal shows outstanding performance when compared to state-of-the-art techniques, not only for huge special events but also for the regular daily traffic.