Computer Science – Artificial Intelligence
Scientific paper
May 2011
adsabs.harvard.edu/cgi-bin/nph-data_query?bibcode=2011aas...21840506w&link_type=abstract
American Astronomical Society, AAS Meeting #218, #405.06; Bulletin of the American Astronomical Society, Vol. 43, 2011
Computer Science
Artificial Intelligence
Scientific paper
AutoClass-C, an artificial intelligence program designed to classify large data sets, was developed by NASA to classify stars based upon their infrared colors. Wanting to investigate its ability to classify asteroidal data, we conducted a preliminary test to determine if it could accurately reproduce the Tholen taxonomy using the data from the Eight Color Asteroid Survey (ECAS). For our initial test, we limited ourselves to those asteroids belonging to S, C, or X classes, and to asteroids with a color difference error of less than +/- 0.05 magnitudes. Of those 406 asteroids, AutoClass was able to confidently classify 85%: identifying the remaining asteroids as belonging to more than one class. Of the 346 asteroids that AutoClass classified, all but 3 (<1%) were classified as they had been in the Tholen classification scheme. Inspired by our initial success, we reran AutoClass, this time including IRAS albedos and limiting the asteroids to those that had also been observed and classified in the Bus taxonomy. Of those 258 objects, AutoClass was able to classify 248 with greater than 75% certainty, and ranked albedo, not color, as the most influential factor. Interestingly, AutoClass consistently put P type objects in with the C class (there were 19 P types and 7 X types mixed in with the other 154 C types), and omitted P types from the group associated with the other X types (which had only one rogue B type in with its other 49 X-types). Autoclass classified the remaining classes with a high accuracy: placing one A and one CU type in with an otherwise perfect S group; placing three P type and one T type in an otherwise perfect D group; and placing the four remaining asteroids (V, A, R, and Q) into a class together.
Campins Humberto
Fernandez Yan R.
Harvell Thomas
Walker Matthew
Ziffer Julie
No associations
LandOfFree
The Potential of AutoClass as an Asteroidal Data Mining Tool does not yet have a rating. At this time, there are no reviews or comments for this scientific paper.
If you have personal experience with The Potential of AutoClass as an Asteroidal Data Mining Tool, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and The Potential of AutoClass as an Asteroidal Data Mining Tool will most certainly appreciate the feedback.
Profile ID: LFWR-SCP-O-1743106