Machine Learning-Based Space Risk Management: Asteroid and Solar Flare Prediction

As humanity continues its exploration of the Universe, safety in space is an ever-growing concern. In this study, we focus on Potentially Hazardous Near-Earth Objects (NEOs), also known as Potentially Hazardous Asteroids (PHAs), that threaten our satellites and Solar Flares that can cause a worldwide blackout of our communication systems. This study combines asteroid hazard and solar flare prediction to enhance space situational awareness and mitigate potential threats through deep learning. We trained our deep learning models using data from diverse sources, including NASA (National Aeronautics and Space Administration), Amerigeoss, Kaggle, and Vizier for NEO detection and classification. The NEO detection mechanism utilizes cascaded binary classification to distinguish between hazardous and non-hazardous NEOs, and the solar flare prediction aspect with data from the DSCOVR (Deep Space Climate Observatory) satellite and Earth's magnetometers, focusing on time-series analysis to forecast geomagnetic storms up to 14 days in advance. Our solar flare prediction model, based on LSTM (Long Short-Term Memory) units, can provide valuable insights for space risk mitigation strategies due to its adaptability to time series data. Integrating asteroid and solar flare prediction capabilities enhances our ability to anticipate and respond to potential hazards in space, thereby contributing to the advancement of space exploration and safety.