Asteroid Hazard Assessment: Optimizing ExtraTrees with RandomizedSearch CV and SHAP Explainable AI

Thisresearch addresses the crucial task of classifying asteroids as hazardous or non-hazardous for effective hazard assessment. Leveraging a dataset obtained from NASA via Kaggle, comprising 4687 rows and 30 columns, we employed various machine learning algorithms to discern the best-suited model. The ExtraTrees classifier emerged as the top performer with an initial accuracy of 96.6%. Notably, further optimization through Randomized Search Cross-Validation elevated the accuracy to 97%. Our study emphasizes the importance of algorithm selection and finetuning in achieving reliable asteroid classification. Motivated by the critical need for accurate hazard assessment, this research highlights the ExtraTree classifier as a robust solution, showcasing its potential in advancing our understanding and mitigation strategies for potential asteroid impacts.