Processed Spatio-Temporal Flood Prediction Dataset for Jakarta

This dataset contains processed temporal, spatial, and graph-based features used to train a hybrid CNN–LSTM–GNN framework for flood risk prediction in Jakarta. The data includes 30-day rainfall sequences, 64-dimensional spatial embeddings derived from DEM, slope, land-use, and drainage-density rasters, as well as concatenated graph embeddings representing hydrological connectivity. Only derived features are provided; raw rainfall, DEM, land-use, and OSM drainage shapefiles are excluded due to redistribution restrictions. The dataset is intended for reproducible research in spatio-temporal deep learning and graph-based environmental modeling. All files are provided in NPZ, NPY, or CSV formats and are compatible with standard Python libraries such as NumPy and pandas. The dataset is released under the CC BY 4.0 license.