Blazars identification among the unidentified sources of the 3FGL gamma-ray catalog

International audienceAfter 4 years of observations, the LAT telescope on board the Fermi satellite has detected the gamma-ray emission from 3034 sources (3FGL catalogue). About one third of these sources have no assigned counterparts at other wavelengths and are of unidentified nature. We present the search of blazar candidates among these unidentified sources. Considering a set of carefully choosen discriminant blazar/non-blazar parameters, we have built two multivariate classifiers (boosted decision trees and multilayer perceptron neural networks) and combined their decisions in order to obtain a high level of blazar indentification efficiency along with a well controlled level of non-blazar contamination. The low latitude and high latitude cases have been treated separately. Applying these classifiers to the sample of 1009 unidentified sources, we have determined a list of 538 blazar candidates, respectively 420 and 118 located at high (|b| > 10◦ ) and low (|b| ≤ 10◦ ) latitude. The corresponding level of mis-classification is estimated to be ∼7 sources for the high latitude case (less than 2%) and ∼28 sources for the low latitude case (∼24%)