It's a pytorch implementation of paper "MagiNet: Mask-Aware Graph Imputation Networkfor Missing Spatio-temporal Data" accepted in TKDD.

pip install -r requirements.txt

1. Download the raw datasets and push them into directory 'datasets/' (download dataset)

   • PEMS-BAY
   • METR-LA
   • Seattle
   • Chengdu
   • Shenzhen

2. Preprocess the dataset:

   python prepare_split_data.py --dataset='METR-LA'
   
   python prepare_miss_data.py --dataset='METR-LA' --miss_mechanism='MCAR' --miss_ratio=0.5 --seqlen=12

python main.py --config_path='configs/METR-LA.yaml' --seed=0