Skip to content

Machine Learning Classification of Potentially Hazardous Asteroids

Notifications You must be signed in to change notification settings

RachelGoodridge/Machine_Learning

Repository files navigation

Potentially Hazardous Asteroids

By: Rachel Goodridge

Abstract

Near-Earth Objects (NEOs), such as asteroids, occasionally have the potential to collide with Earth and cause enormous damage. NASA has been documenting these NEOs as well as many other types of space-going debris and developing diversion tactics given the event of an impact threat. From NASA’s Small-Body Database, information about asteroids was used to predict which are potentially hazardous. Following oversampling to address class imbalance and optimization of many parameters, the best model based on F1 validation scores was XGBoost Classifier. The score on the holdout data was 0.784 and the most important features were absolute magnitude, closest distance to the sun, and orbit uncertainty.

About

Machine Learning Classification of Potentially Hazardous Asteroids

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published