#NTUST-DM-Final_project-reivsed by (https://github.com/RobinLu1209/ST-GFSL#spatio-temporal-graph-few-shot-learning-with-cross-city-knowledge-transfer)

Spatio-Temporal Graph Few-Shot Learning with Cross-City Knowledge Transfer

Requirements

• torch >= 1.8.1
• numpy >= 1.20.3
• scikit-learn >= 0.24.2
• pytorch geometric >= 1.7.2
• pyaml
• scipy
• tqdm

Data

• Download the revised_dataset which added pems04 and pems08 by google drive

Model training & Testing(How to implement the code)

# To the floder that main.py was placed 
cd ./ST-GFSL-Custom_Dataset
# Set the test dataset and model structure
CUDA_VISIBLE_DEVICES=[CUDA NUM] python main.py --test_dataset [FEW-SHOT DATASET] --model [ST-META MODEL]
# For example: Use GRU model to train a model, and test on PEMS-BAY datasets
CUDA_VISIBLE_DEVICES=0 python main.py --test_dataset pems-04 --model GRU

Citation

@inproceedings{DBLP:conf/KDD/CrossCityTransfer22,
  author    = {Bin Lu and
               Xiaoying Gan and
               Weinan Zhang and
               Huaxiu Yao and
               Luoyi Fu and
               Xinbing Wang},
  title     = {Spatio-Temporal Graph Few-Shot Learning with Cross-City Knowledge Transfer},
  booktitle = {{KDD} '22: The 28th {ACM} SIGKDD Conference on Knowledge Discovery and Data Mining,
              Washington, DC, USA, August 14--18, 2022},
  publisher = {{ACM}},
  year      = {2022}
}