Predicting Potentially Hazardous Asteroids (PHAs)

Predicting Potentially Hazardous Asteroids (PHAs) helps mitigate the risk of their possible impacts on Earth, which is one of the emerging data science applications. In this case study, data science techniques are applied to improve the accuracy of PHA predictions. The objective of this research is to classify asteroids by their risk level, which requires a dataset that contains all the necessary orbital and physical parameters and is collected from the available astronomical data. Key features used from the datasets include eccentricity, inclination, semi-major axis, as well as physical attributes such as absolute magnitude, size, and albedo. To handle missing values for the more prominent asteroid characteristics, like diameter, mathematical imputation methods are implemented. The inclusion of machine learning techniques like Random Forest with existing traditional orbital mechanics increases the effectiveness of the long-term impact estimation and solves some of the limitations inherent in the accepted methodologies. This research seeks to reduce the incorrect classification of PHAs by improving the general accuracy of the classification using a multitude of features, such as orbital dynamics, physical features, and uncertainty estimation.