Security checking is a major issue in airport operations. Affecting the correct number of security
agents is essential to provide a good quality of service to passengers while providing the best security
performances. At Paris Charles de Gaulle airport the affectation of security agents is decided at strategical
level, more than a month in advance. The key element to determine the number of agents needed is the
passenger flow through the security checkpoints. This flow is correlated to the passenger flow in the
different boarding rooms. This paper investigates the interest of small dense neural networks to perform
passenger flow prediction at strategical level for Paris Charles de Gaulle airport. A dense neural network
has been trained to predict the passenger flow for each boarding room of the airport. The network has
been compared to a more complex long short-term memory model in terms of mean absolute error and
outperformed a mathematical model based on exponentially modified Gaussian distribution.