TOWARD A METEORITE-BASED ASTEROID TAXONOMY
Meteorite classification depends on mineralogy, petrology, chemistry, and O-isotopes. Because meteorites can give higher resolution data and are easier to study, the classification process for meteorites is more robust than that of asteroids. We are using 2401 spectra of the well-defined meteorites as a starting point for a new asteroid taxonomy, leveraging >700 asteroid spectra from various sources. All meteorite and asteroid spectra were resampled to cover the range from 0.3 to 2.5 µm at 0.5 nm resolution and relabeled with a common set of headers for ease of analysis. Spectra were classified using logistic regression and k-nearest neighbors, two common machine learning algorithms. Results show that most meteorite classes have distinct spectra that are readily recognized by classifiers. For example, Nakhlites can be classified with 100% accuracy and CI meteorites can be classified with 89% accuracy. Because these groups’ spectra are so distinctive, we can match any individual asteroid spectrum against meteorite/asteroid data quantitatively. We are also experimenting with classifying the entire database (meteorites and asteroids) together to assess similarities among objects and groups of objects. The end result will be to develop a modern classification algorithm that extends well-characterized meteorite classes to a new asteroid taxonomy.