This work proposes a probabilistic decision tool that integrates the main factors influencing food insecurity in Brazil. Food security exists when individuals have access to sufficient food to meet their dietary needs in quantity and quality. Evidence indicates that food insecurity is associated with malnutrition, diabetes, cardiovascular diseases, some cancers, deterioration of mental health, inability to manage chronic disease, worse children's academic performance, and social skills. Government policies on welfare, farming, the environment, employment, health, and others impact food security at various levels. Each of these is a dynamic sub-system that considers factors influencing food insecurity such as weather, economy, food availability, social and demographic aspects. In the context of policies for complex systems, it is difficult for decision-makers to account for all the variables within the system. The usual approach to relate factors and outcomes is based on linear regression models that do not allow for cause-effect inference. Our proposal is based on Bayesian networks that can capture both non-linearities and complex cause-effect relationships. This model considers a time-varying Dirichlet process for smooth changes in the economic, demographic, and weather series affecting food security. Besides, it is able to handle data from different surveys, unevenly spaced time series, and missing/non-response data. Thus, the effectiveness and sustainability of interventionist policies such as the \textit{Fome Zero} program -- the larger initiative to combat hunger in Brazil, can be measured. Food security was measured using a national psychometric scale consisting of questions related to the direct experience of food insecurity available on the Brazilian National Household Sample Survey (PNAD). The outcome of this project is a decision support system that integrates the main factors influencing food insecurity in a probabilistic model. With this tool, the decision-makers will optimize the cost-effectiveness of future interventions, simulate scenarios, and compare several candidate policies.