^1,2M. Darby Dyar,¹Sydney M. Wallace,^1,2Thomas H. Burbine,³Daniel R. Sheldon
Icarus (in Press) Link to Article [https://doi.org/10.1016/j.icarus.2023.115718]
¹Planetary Science Institute, 1700 East Fort Lowell, Suite 106, Tucson, AZ 85719-2395, USA
²Department of Astronomy, Mount Holyoke College, 50 College St., South Hadley, MA 01075, USA
³College of Information and Computer Sciences | 140 Governors Dr., Amherst, MA 01003, USA
Copyright Elsevier

Understanding the distribution of matter within our Solar System requires a robust methodology for evaluating the composition of small objects in the asteroid belt. Existing asteroid taxonomies have variously been based on spectral features relating to mineralogy and on classification of asteroid spectra alone. This project tests a fundamentally different approach, using machine learning algorithms to classify asteroids based on spectroscopic characteristics of existing meteorite classes. After evaluating four classification techniques built on labeled meteorite spectral data, logistic regression (LR) was determined to provide the most accurate results that distinguish eight robust groups of meteorite classes to which asteroid spectra can then be matched. The groups are rooted in mineralogical composition and directly relate meteorites to potential host bodies. A standalone LR algorithm classifies unknown asteroid spectra uniquely as one of eight specific group, allowing the distribution of compositions in the asteroid belt to be evaluated.