An Empirical Experiment on Deep Learning Models for Predicting Traffic Data

This repository contains detailed context and experimental results of the paper "An Empirical Experiment on Deep Learning Models for Predicting Traffic Data," ICDE 2021.

Experiment Settings

In the experiment, we evaluated 8 models including STGCN, DCRNN, ASTGCN, ST-MetaNet, Graph-WaveNet, STG2Seq, STSGCN, and GMAN, with 7 datasets, including METR-LA, PeMS-BAY, PeMSD7(M), PeMSD3, PeMSD4, PeMSD7, and PeMSD8. For the fair evaluation, we implemented all the models into PyTorch and process the input with same manner, such as sequence length.

Additional Information including source code will be updated.