This thesis work consists in studying Deep Learning models in order to predict, as well as possible, the Internet traffic in a mobile network environment and use them in the 5G context for MEC (Multi-access Edge Computing) dimensioning. In particular, the first chapter gives a brief introduction about AI and Machine Learning. Basic concepts needed to understand the mechanism behind Deep Learning are provided through the second chapter. Third chapter supplies a description of used dataset. Analysis and results of DL models are presented in chapter four. Finally, the last chapter shows an application of the whole study for MEC dimensioning.