carrtesy / Graph-WaveNet-WaveBound

Exemplary Implementation of WaveBound (NeurIPS 2022) to Traffic domain, especially GWNet.

WaveBound + GraphWaveNet

Implementation of WaveBound: Dynamic Error Bounds for Stable Time Series Forecasting (NeurIPS 2022), applied to Graph WaveNet for Deep Spatial-Temporal Graph Modeling, IJCAI 2019. This repo is officially provided as examplary code for applying WaveBound method to Traffic domain. To view WaveBound's official implementation, visit the wavebound-github-repo.

Data Preparation

Step1: Download METR-LA and PEMS-BAY data from Google Drive or Baidu Yun links provided by DCRNN.

Step2: Process raw data

# Create data directories
mkdir -p data/{METR-LA,PEMS-BAY}

# METR-LA
python generate_training_data.py --output_dir=data/METR-LA --traffic_df_filename=data/metr-la.h5

# PEMS-BAY
python generate_training_data.py --output_dir=data/PEMS-BAY --traffic_df_filename=data/pems-bay.h5

Train Commands

sh scripts/metr_la.sh # METR-LA
sh scripts/pems_bay.sh # PEMS-BAY

Quantitative Results

METR-LA

STEPS	15min			30min			60min
Metrics	MAE	RMSE	MAPE	MAE	RMSE	MAPE	MAE	RMSE	MAPE
GWNet	2.6998	5.1684	0.0693	3.0966	6.2413	0.0837	3.5812	7.4361	0.1010
GWNet+WaveBound	2.6703	5.1104	0.0691	3.0313	6.1046	0.0832	3.4478	7.1474	0.0990

PEMS-BAY

STEPS	15min			30min			60min
Metrics	MAE	RMSE	MAPE	MAE	RMSE	MAPE	MAE	RMSE	MAPE
GWNet	1.3003	2.7225	0.0273	1.6171	3.6546	0.0364	1.9201	4.4375	0.0449
GWNet+WaveBound	1.2917	2.7221	0.0269	1.6080	3.6538	0.0359	1.9045	4.4103	0.0449

Qualitative Results