Hybrid Approach for Accurate Asteroid Path Prediction and Classification Using Ensemble Learning

This study presents a novel hybrid methodology that integrates classical physics-based orbital propagation with machine learning (ML) corrections to forecast the future trajectories of near-Earth asteroids while also assessing their hazard potential. The approach combines an analytical two-body orbit model with ML-based error correction—implemented using Long Short-Term Memory (LSTM) network—to significantly reduce prediction errors. Additionally, ensemble learning techniques using XGBoost and Random Forest are employed for asteroid hazard classification, achieving an accuracy of 99.08%. The results demonstrate that the hybrid approach not only improves long-term trajectory prediction accuracy but also enhances risk assessment, paving the way for more reliable planetary defense and space mission planning. Key Words: Asteroid trajectory prediction, orbital mechanics, hybrid model, machine learning, neural networks, LSTM, XGBoost, Random Forest, ensemble classification, hazard assessment