Abstract—MPLS-TE Networks are complex interacting systems, providers integrate absolutely the forwarding equivalent class involving in the dynamic routing algorithm "Dynamic Multi Criteria Routing", especially with network services such as an IPtv, a VPNs or a VoIP with the several types of video streaming and similar user applications, making its flagship technologies of tomorrow. Machine learning is very compatible for this kind of networks. In this paper, we first review how the main machine learning concepts can be integrated in communication networks and discuss concepts, supervised and unsupervised model suitable to the network, data and management strategies, and creating a new architecture of network controls and management tools. We then describe case studies networking in detail, anticipate anomalies at multiple network layers, covering predictive maintenance, descriptive network topology management, capacity optimization. Finally, we prove the importance of this work, and guess an overview of intelligent dynamic networks.