With the raising of social networks and social sensing, several methods and applications focused on urban and economic analysis emerged to assist urban planning and business solutions. Thus, the inhabitants of a city share social data according to their context, acting as sensors. In this way, location-based social networks have become important because they associate social data with precise locations, which allows promising studies. In this work, we present a method of popularity analysis in urban regions through Foursquare, using the concept of functional regions. For this, we used 204,737 check-ins for 2 years from the same locality for a case study. In this way, we consider not only the main functions of each region, but also the secondary functions. We also present a machine learning model in which a neural network is used to define strategic regions for venues considering not only the subcategories of Foursquare, but also the macrocategories.