Nowadays the equitable distribution of media subsidies in the world represents one of the greatest challenges. In Colombia, the government seeks to focus public and social spending on the poorest and most vulnerable population in the territory. To achieve this goal, a Beneficiaries Selection System for Social Programs (SISBEN) has been created to obtain socio-economic information from population groups. It is used to determine the possible benefits of government subsidies in the areas of health, education, among others. However, the system has irregularities at the time of selection, which causes the budget to be granted to people who do not deserve it. In the present article, a neural model is proposed for the classification of the possible beneficiaries of the SISBEN through a mathematical method that groups by centroids the observations, which allows grouping the applicants according to their characteristics in clusters.