Flow prediction research has attracted considerable interest as it has the potential to provide fundamental support for addressing several critical challenges in the field of public safety and smart cities, specifically, predicting flows of both people and vehicles to aid in the efficient organization of resources (e.g., healthcare or public order). The present study seeks the ability to identify various events that exhibit distinctive characteristics associated with potentially dangerous situations in large-scale crowds of people. An analysis of current research shows that there are still situations to be defined, especially linked to the concept of early warning and anomaly detection, the timing of which could be key in the outcome of a possible risk event.