|
|
--- |
|
|
tags: |
|
|
- traffic-forecasting |
|
|
- time-series |
|
|
- graph-neural-network |
|
|
- graph-wavenet |
|
|
datasets: |
|
|
- metr-la |
|
|
--- |
|
|
|
|
|
# Graph-WaveNet Model - METR-LA |
|
|
|
|
|
Graph WaveNet for traffic speed forecasting, combining graph convolution with dilated causal convolution. |
|
|
|
|
|
## Model Description |
|
|
|
|
|
This model uses a graph neural network architecture that combines: |
|
|
- Adaptive adjacency matrix learning |
|
|
- Spatial graph convolution for capturing spatial dependencies |
|
|
- Temporal convolution with dilated causal convolutions |
|
|
- Multi-scale temporal receptive field |
|
|
|
|
|
## Evaluation Metrics |
|
|
|
|
|
- **Test MAE (15 min)**: 2.4767 |
|
|
- **Test MAPE (15 min)**: 0.0618 |
|
|
- **Test RMSE (15 min)**: 4.5875 |
|
|
|
|
|
|
|
|
## Dataset |
|
|
|
|
|
**METR-LA**: Traffic speed data from highway sensors. |
|
|
|
|
|
## Usage |
|
|
|
|
|
```python |
|
|
from utils.gwnet import load_from_hub |
|
|
|
|
|
# Load model from Hub |
|
|
model = load_from_hub("METR-LA") |
|
|
|
|
|
# Get predictions |
|
|
import numpy as np |
|
|
x = np.random.randn(10, 12, 207, 2) # (batch, seq_len, nodes, features) |
|
|
predictions = model.predict(x) |
|
|
``` |
|
|
|
|
|
## Training |
|
|
|
|
|
Model was trained using the Graph-WaveNet implementation with default hyperparameters. |
|
|
|
|
|
## Citation |
|
|
|
|
|
If you use this model, please cite the original Graph WaveNet paper: |
|
|
|
|
|
```bibtex |
|
|
@inproceedings{wu2019graph, |
|
|
title={Graph WaveNet for Deep Spatial-Temporal Graph Modeling}, |
|
|
author={Wu, Zonghan and Pan, Shirui and Long, Guodong and Jiang, Jing and Zhang, Chengqi}, |
|
|
booktitle={Proceedings of the 28th International Joint Conference on Artificial Intelligence}, |
|
|
pages={1907--1913}, |
|
|
year={2019} |
|
|
} |
|
|
``` |
|
|
|
|
|
## License |
|
|
|
|
|
This model checkpoint is released under the same license as the training code. |
|
|
|
