With the expansion of educational programs, observed in the exponential increase in access to tertiary education, views have arisen that claim that it is not only necessary to implement policies to widen access coverage, but also to emphasize aspects related to the quality of education, such as dropout and retention. This research seeks to establish a model of analysis to promote retention and prevention of the causes involved in dropout. The supervised learning methodology is used to determine the variables of analysis, through regression techniques and decision trees, generating a logistic model of desertion, capable of improving the efficiency in the allocation of resources. It is expected to contribute to the understanding of phenomena associated with university dropout and retention and to contribute to the management and decision-making process of higher education institutions.