3D Graph Convolutional Networks with Temporal Graphs: A Spatial Information Free Framework For Traffic Forecasting

TL;DR

This paper introduces 3D-TGCN, a novel traffic forecasting framework that learns spatial relationships from time series similarity and employs 3D graph convolutions to better capture spatio-temporal dynamics, outperforming existing methods.

Contribution

The paper presents a spatial information free approach to traffic prediction using learned graph structures and a new 3D graph convolution model, improving accuracy and efficiency.

Findings

Outperforms state-of-the-art baselines in traffic forecasting.

Learns spatial relationships from time series similarity.

Uses 3D graph convolutions for better spatio-temporal modeling.

Abstract

Spatio-temporal prediction plays an important role in many application areas especially in traffic domain. However, due to complicated spatio-temporal dependency and high non-linear dynamics in road networks, traffic prediction task is still challenging. Existing works either exhibit heavy training cost or fail to accurately capture the spatio-temporal patterns, also ignore the correlation between distant roads that share the similar patterns. In this paper, we propose a novel deep learning framework to overcome these issues: 3D Temporal Graph Convolutional Networks (3D-TGCN). Two novel components of our model are introduced. (1) Instead of constructing the road graph based on spatial information, we learn it by comparing the similarity between time series for each road, thus providing a spatial information free framework. (2) We propose an original 3D graph convolution model to model the spatio-temporal data more accurately. Empirical results show that 3D-TGCN could outperform state-of-the-art baselines.