Uma metodologia para avaliar modelos de previsão de eventos a partir de redes sociais

Social networks have been used as data sources to predict the occurrence of real events. However, the success of such predictions depends on the ability of the model to capture the characteristics of the event in focus, and determining the most appropriate models is a challenge. In this work we present a methodology to evaluate comparatively different models with a view to selecting the most appropriate one. We also apply our methodology in the context of predicting the dengue epidemic, being able to identify the best model for each city to be employed and the characteristics that justify its choice.