The 5G promise for ubiquitous communications is expected to be a key enabler for transportation efficiency. However, the consequent increase of both data payload and number of users derived from new Intelligent Transport Systems makes network management even more challenging; an ideal network management will need to be capable of self-managing fast-moving nodes that sit in the 5G data plane. Platooning applications, for instance need a highly flexible and high efficient infrastructure for optimal road capacity. Network management solutions have, then, to accommodate more intelligence in its decision-making process due to the network complexity of ITS. This study proposes this envisioned architecture, namely cognition-enabled network management, for 5G V2X communication (CogITS). It is empowered by machine learning to dynamically allocate resources in the network based on traffic prediction and adaptable physical layer settings. Preliminary proof-of-concept validation results, in a platooning scenario, show that the proposed architecture can improve the overall network latency over time with a minimum increase of control message traffic.