A disaster is an event that directly causes social, economic, and/or environmental impacts on a society. Floods, droughts, and infectious diseases are examples of disasters. The difficulty in anticipating the extent of these events is a barrier for municipal, state, and federal governments to organize strategies in advance to contain the effects caused. In this sense, we can mention the creation of welfare policies that aim to assist individuals and families financially. To this end, it is important to understand how the occurrence of a disaster and the declaration of an emergency or calamity by municipalities are related to the increase or creation of such policies, as well as the dynamics of their implementation and impacts on the affected population. Machine Learning (ML) offers means to attack problems similar to those mentioned above by considering approaches such as pattern discovery and predictions, among others. The objective of this work is to analyze the relationships between disaster data and social indicators through the application of ML algorithms that generate predictive models. To this end, data from the Cadastro Único (CadÚnico), a database of information on low-income families, and Bolsa Família, a social assistance program created in 2003 to transfer a monthly income to low-income families, were considered. As a result, the ML models used are presented and described, and a comparative analysis of their performance is performed, in addition to the interpretation of the patterns revealed that may be used as a basis for decisions by the actors involved.