$\text{Explainable Deep Learning Based Potentially Hazardous Asteroids Classification}$ $\text{Using Graph Neural Networks}$

This repository implements a Graphical Neural Networks and its variants (Graph Attention Networks and GraphSAGE) models to classifiy potentially hazardous asteroids on the NASA Jet Propulsion Lab's Small Body Datasets.

$\text{●•Authors}$

$\text{Baimam Boukar Jean Jacques}$

Carnegie Mellon University Africa

bbaimamb@andrew.cmu.edu

●• $\text{Dataset Description}$

Our dataset, the Asteroid Dataset, is from NASA's Jet Propulsion Laboratory (JPL). It contains over 950,000 records, sourced from the official Small-Body Database by the NASA Jet Propulsion Lab. It was originally preprocessed by a NASA Astronomy and Astrophysics Researcher.

The preprocessed version is publicly available, and licensed under OpenData Commons Open Database License (ODbL) v1.0 by a JPL-authored document sponsored by NASA under Contract NAS7-030010.

The dataset contains detailed information on thousands of asteroids. Its main attributes include orbital eccentricity, semimajor axis,perihelion distance, absolute magnitude, diameter, and the Near-Earth Object (NEO) and Potentially Hazardous Asteroid (PHA) flags.

●• $\text{Methodology}$

●• $\text{Reproduction Steps}$

●• $\text{Cite This Paper}$

@software{baimamboukar_2025,
author = {Baimam Boukar Jean Jacques},
month = apr,
title = {{Explainable Deep Learning Based Potentially Hazardous Asteroids Classification Using Graph Neural Networks}},
url = {https://github.com/baimamboukar/hazardous-asteroid-classification},
version = {1.0},
year = {2025}
}