Skip to main content
All Reviews
Astronomy
intermediate

Machine learning classification of new asteroid families members

Carruba, V. et al. (2020)

Published
May 28, 2020
Journal
Monthly Notices of the Royal Astronomical Society · Vol. 496 · No. 1
DOI
10.1093/mnras/staa1463

At a GlanceAI

ML classifiers (esp. ExtraTrees) rapidly recover ~97% of HCM asteroid-family members from proper-element distributions.

SummaryAI

The study tackles the growing challenge of assigning newly discovered asteroids to collisional families as catalogues expand beyond what traditional clustering can easily handle. Using the proper-element distribution (a, e, sin i) of known family members as training data, the authors compare nine machine-learning classifiers to predict additional members. Extremely randomized trees (ExtraTrees) achieves the best performance, recovering up to 97% of the members identified by the standard hierarchical clustering method. The result suggests ML can serve as a scalable, automated complement to HCM for updating family memberships as surveys grow.

Method SnapshotAI

Supervised machine-learning classification in proper-element space, with a comparative benchmark of multiple algorithms against HCM labels.

BackgroundAI

Basic understanding of asteroid families, proper orbital elements, and standard classification/clustering concepts.