Presentation #103.33 in the session Missions and Instruments.
The Chandra Source Catalog (CSC) is the ultimate repository of sources detected by Chandra and represents a fertile ground for discovery in high energy astrophysics. A significant fraction of the sources in CSC have not been studied in detail and only a small fraction of them have been classified. Among the potentially interesting sources that we can look for in Chandra data are compact object mergers, extrasolar planet transits, tidal disruption events, etc. In order to conduct a thorough investigation of the CSC sources, we need to classify them as far as we can. This work proposes an unsupervised machine learning approach to classify previously uncategorized Chandra Source Catalog sources by using only the X-ray data available. We propose a new methodology that performs Gaussian Mixture clustering to a set of CSC source properties, and then associates cluster members with objects previously classified in order to provide a probabilistic classification for the sources. Together with the catalog, we also provide visualization tools that allow us to visualize the classes projected on the sky, and study the astrophysical implications of our result. We demonstrate that under certain circumstances, we can distinguish between X-ray binaries from AGNs candidates. We classify YSOs with certain confidence in the galactic plane, which in particular cases is confused depending on the level of obscuration. Overall, our results demonstrate that it is possible to assign probabilistic classes to a significant fraction of new X-ray sources based on the X-ray information only.