Title

A New Tool for Classifying Small Solar System Objects

Authors

Ryan Desfosses, University of Southern Maine
Derick Arel, University of Southern Maine
Matthew E. Walker PhD, University of Southern MaineFollow
Julie Ziffer PhD, University of Southern MaineFollow
Thomas Harvell, University of Southern Maine
Humberto Campins PhD, University of Central FloridaFollow
Yanga R. Fernández PhD, University of Central FloridaFollow

Document Type

Conference Proceeding

Publication Date

5-2011

Publication Title

American Astronomical Society, Department of Planetary Sciences

Abstract

An artificial intelligence program, AutoClass, which was developed by NASA's Artificial Intelligence Branch, uses Bayesian classification theory to automatically choose the most probable classification distribution to describe a dataset. To investigate its usefulness to the Planetary Science community, we tested its ability to reproduce the taxonomic classes as defined by Tholen and Barucci (1989). Of the 406 asteroids from the Eight Color Asteroid Survey (ECAS) we chose for our test, 346 were firmly classified and all but 3 (<1%) were classified by Autoclass as they had been in the previous classification system (Walker et al., 2011). We are now applying it to larger datasets to improve the taxonomy of currently unclassified objects. Having demonstrated AutoClass's ability to recreate existing classification effectively, we extended this work to investigations of albedo-based classification systems. To determine how predictive albedo can be, we used data from the Infrared Astronomical Satellite (IRAS) database in conjunction with the large Sloan Digital Sky Survey (SDSS), which contains color and position data for over 200,000 classified and unclassified asteroids (Ivesic et al., 2001). To judge our success we compared our results with a similar approach to classifying objects using IRAS albedo and asteroid color by Tedesco et al. (1989). Understanding the distribution of the taxonomic classes is important to understanding the history and evolution of our Solar System. AutoClass's success in categorizing ECAS, IRAS and SDSS asteroidal data highlights its potential to scan large domains for natural classes in small solar system objects. Based upon our AutoClass results, we intend to make testable predictions about asteroids observed with the Wide-field Infrared Survey Explorer (WISE).

Recommended Citation

Desfosses, Ryan & Arel, Derick & Walker, Matthew & Ziffer, Julie & Harvell, Thomas & Campins, Humberto & Fernandez, Yanga. (2011). "A New Tool for Classifying Small Solar System Objects." AAS/Division for Planetary Sciences Meeting #218.