Cold Wave Dataset

This dataset supports the research and code presented in the paper, Dynamic Spatio-Temporal Graph Neural Network for Cold Wave Prediction. The uploaded ZIP file contains all raw and processed data necessary to reproduce the findings in the associated GitHub repository. Data Contents The data is organized into three main directory structures: 1. `raw/`: Contains the original, publicly available weather station data, including 15 individual CSV files (`data1.csv` through `data15.csv`) and external long-term temperature mean NetCDF files from the `Tmin_longterm_mean` directory.2. `datasets/`: Contains the cleaned and stabilized data splits used directly by the modeling scripts, including `part_1.csv` (5 years data of 51 weather stations) and `part_2.csv` (10 years data of 115 weather stations), and the combined `DATASET.csv`.3. `processed/`: Contains all the graph dataset variations obtained by running graph_formulation.py file on the two parts of the dataset. Code and Reproducibility The code used for preprocessing, graph generation, model training, and evaluation is available in the linked GitHub repository. The feature names, scaling methods, and graph construction parameters for this data are defined in the repository's `config.yaml` and documented in the `README.md`. Code Repository (GitHub): Repo Link Final Repository (Zenodo): 10.5281/zenodo.17680585