Searching for open clusters with density-based clustering algorithms in Gaia era

Abstract

The imminent publication of the second Gaia data release (GDR2) scheduled for April2018 will mark the beginning of Big Data in Astrometry. On the one hand, this will require an adaptation of IT infrastructures and software to cope with the huge volumeand diversity of data. On the other hand, by making available the positions, velocities, and photometric measurements for more than one billion stars with unprecedentedaccuracy, it paves the way for Machine Learning and Data Mining techniques to be used in various areas of Galactic Astronomy. In this context, the search for Galactic open clusters is probably one of the fieldsof research which will benefit most from this new data. Indeed, owing to their history, these groups of stars naturally share many properties, and are thus very well suitedfor the use of clustering algorithms. This work investigates the potential of densitybased algorithms (kNN, DBSCAN, OPTICS) in the search for open clusters withina 5D astrometric space. It proposes a method to carefully set hyperparameters andapplies it to the analysis of Tycho-Gaia Astrometric Solution (TGAS) data, yielding 60 new open cluster candidates. This algorithm will eventually be scalable to GDR2 and other forthcoming large astrometric datasets.