This paper analyses how to build an efficient classifier across Bayesians networks used in data mining. The purpose of using the Bayesian model is to improve credit scoring accuracy. The Bayesian approach, based on probability models, analyses risk by using the decision theory, yielding as a solution that action that maximizes the expected utility. Expert assessment may be included in the model. To show the superiority of the Bayesian approach, results obtained for real bank data are compared with those obtained with alternative parametric and non-parametric models. © 2013 Asociación Cuadernos de Economía.