This is the code for paper:
ST-Imputer: Context-aware Spatiotemporal Imputation Diffusion Network with Physics Guidance
Recent versions of the following packages for Python 3 are required:
- numpy==1.24.1
- PyYAML==6.0.1
- Requests==2.31.0
- scikit_learn==1.5.0
- scipy==1.13.1
- setuptools==69.5.1
- tensorboardX==2.6.2
- torch==1.13.1
- torch_cluster==1.6.0
- torch_geometric==2.5.3
- torch_sparse==0.6.15
- tqdm==4.66.4
All of the datasets we use are publicly available datasets.
The used datasets are available at:
- GLODAPv2 https://glodap.info/
- ECMWF https://cds.climate.copernicus.eu/
- CMEMS https://marine.copernicus.eu/
- AQI-36 http://research.microsoft.com/apps/pubs/?id=264768
Use the following command to run the main script:
python aqi_main.py
If you want to modify the experimental configuration parameters, please refer to /config/base.yaml and config.py. Specifically, key parameters such as the epochs, feature dimensions, and the number of dataset nodes can all be modified in the configuration file.
Convert the data file into .txt format, place it in the \data directory, update the corresponding path in the DA_AQI file, and modify the num_nodes in config.py.