Approximately one-third of the gamma-ray sources in the third Fermi-LAT catalog are unidentified or unassociated with objects at other wavelengths. Observations with the X-Ray Telescope on the Neil Gehrels Swift Observatory (Swift-XRT) have yielded possible counterparts in ~30% of these source regions. The objective of this work is to identify the nature of these possible counterparts, utilizing their gamma-ray properties coupled with the Swift derived X-ray properties. The majority of the known sources in the Fermi catalogs are blazars, which constitute the bulk of the extragalactic gamma-ray source population. The galactic population on the other hand is dominated by pulsars. Overall, these two categories constitute the majority of all gamma-ray objects. Blazars and pulsars occupy different parameter space when X-ray fluxes are compared with various gamma-ray properties. In this work, we utilize the X-ray observations performed with the Swift-XRT for the unknown Fermi sources and compare their X-ray and gamma-ray properties to differentiate between the two source classes. We employ two machine-learning algorithms, decision tree and random forest (RF) classifier, to our high signal-to-noise ratio sample of 217 sources, each of which corresponds to Fermi unassociated regions. The accuracy scores for both methods were found to be 97% and 99%, respectively. The RF classifier, which is based on the application of a multitude of decision trees, associated a probability value (P_bzr_) for each source to be a blazar. This yielded 173 blazar candidates from this source sample, with P_bzr_>=90% for each of these sources, and 134 of these possible blazar source associations had P_bzr_>=99%. The results yielded 13 sources with P_bzr_<=10%, which we deemed as reasonable candidates for pulsars, seven of which result with P_bzr_<=1%. There were 31 sources that exhibited intermediate probabilities and were termed ambiguous due to their unclear characterization as a pulsar or a blazar.
Cone search capability for table J/ApJ/887/18/table1 (Classification with machine learning)
