Home

Awesome

TwoResNet

<picture> <source media="(prefers-color-scheme: dark)" srcset="./figs/tworesnet_dark.svg"> <source media="(prefers-color-scheme: light)" srcset="./figs/tworesnet_light.svg"> <img alt="Shows an illustrated sun in light color mode and a moon with stars in dark color mode." src="https://user-images.githubusercontent.com/25423296/163456779-a8556205-d0a5-45e2-ac17-42d089e3c3f8.png"> </picture>

This is a PyTorch lightning implementation of Two-Level resolution Neural Network (TwoResNet) for traffic forecasting.

1. Installing dependencies

1.1. Create a venv environment

python3 -m venv env

1.2. Activate pip environment

source env/bin/activate

1.3. Install dependencies

pip install -r requirements.txt

1.4. Install PyTorch

In case error occurs, try to install PyTorch according to your local environment following the description here.

2. Model training

You can find tensorboard logs for pretrained models here.

2.1. METR-LA

python run.py --config=data/config/training.yaml --train --dataset=la

2.2. PEMS-BAY

python run.py --config=data/config/training.yaml --train --dataset=bay

3. Test

3.1. METR-LA

python run.py --config=data/config/test.yaml --test --dataset=la

Result

Horizon 1 (5 min) - MAE: 2.24, RMSE: 3.86, MAPE: 5.32
Horizon 2 (10 min) - MAE: 2.49, RMSE: 4.60, MAPE: 6.19
Horizon 3 (15 min) - MAE: 2.65, RMSE: 5.08, MAPE: 6.78
Horizon 4 (20 min) - MAE: 2.79, RMSE: 5.47, MAPE: 7.29
Horizon 5 (25 min) - MAE: 2.90, RMSE: 5.79, MAPE: 7.73
Horizon 6 (30 min) - MAE: 3.01, RMSE: 6.07, MAPE: 8.14
Horizon 7 (35 min) - MAE: 3.09, RMSE: 6.30, MAPE: 8.47
Horizon 8 (40 min) - MAE: 3.17, RMSE: 6.51, MAPE: 8.78
Horizon 9 (45 min) - MAE: 3.23, RMSE: 6.68, MAPE: 9.05
Horizon 10 (50 min) - MAE: 3.29, RMSE: 6.83, MAPE: 9.28
Horizon 11 (55 min) - MAE: 3.34, RMSE: 6.96, MAPE: 9.50
Horizon 12 (60 min) - MAE: 3.39, RMSE: 7.08, MAPE: 9.71
Aggregation - MAE: 2.97, RMSE: 6.01, MAPE: 8.02

3.2. PEMS-BAY

python run.py --config=data/config/test.yaml --test --dataset=bay

Result

Horizon 1 (5 min) - MAE: 0.87, RMSE: 1.56, MAPE: 1.67
Horizon 2 (10 min) - MAE: 1.12, RMSE: 2.21, MAPE: 2.26
Horizon 3 (15 min) - MAE: 1.30, RMSE: 2.73, MAPE: 2.72
Horizon 4 (20 min) - MAE: 1.43, RMSE: 3.14, MAPE: 3.08
Horizon 5 (25 min) - MAE: 1.53, RMSE: 3.45, MAPE: 3.37
Horizon 6 (30 min) - MAE: 1.61, RMSE: 3.69, MAPE: 3.60
Horizon 7 (35 min) - MAE: 1.68, RMSE: 3.88, MAPE: 3.79
Horizon 8 (40 min) - MAE: 1.73, RMSE: 4.03, MAPE: 3.95
Horizon 9 (45 min) - MAE: 1.78, RMSE: 4.15, MAPE: 4.09
Horizon 10 (50 min) - MAE: 1.82, RMSE: 4.25, MAPE: 4.21
Horizon 11 (55 min) - MAE: 1.85, RMSE: 4.33, MAPE: 4.31
Horizon 12 (60 min) - MAE: 1.89, RMSE: 4.41, MAPE: 4.40
Aggregation - MAE: 1.55, RMSE: 3.59, MAPE: 3.45

4. Citation

If you find this repository, e.g., the code and the datasets, useful in your research, please cite the following paper:

@inproceedings{Li2022tworesnet,
      title = {TwoResNet: Two-level resolution neural network for traffic forecasting of freeway networks},
      author = {Li, Danya and Kwak, Semin and Geroliminis, Nikolas},
      year = {2022},
      publisher={25th IEEE International Conference on Intelligent Transportation Systems (ITSC)},
      venue = {Macau, China}, eventdate={2022-10-08/2022-10-12},
}