We present a classification algorithm based on bayesian probabilities computed from the Besançon model of the Galaxy. The scheme is as follows: A simulation in the direction of the stars to be classified is performed from the Galaxy model. Then, assuming values and errors on the observables (e.g., G magnitude, colours, kinematics,...), probabilities of a star to belong to a given stellar population , and to have such luminosity class, spectral type, and other intrinsic parameters, are deduced,

more »

... ased on the probabilities of having such combination of observables in the simulation. This method can be used further to classify Gaia stars during the mission. It can also be used to identify stars having unexpected observables measured by Gaia instruments and to trigger a Science Alert.