Future 5G infrastructures promise to deliver un- precedented Quality of Services (QoS) guarantees through ultra- low latency and high data rate specifications that lead to handle many critical applications and services. Network Function Virtualization (NFV) is the paradigm that enables the implementation of network functions and capabilities as software components executed in virtual entities provisioned in general-purpose hardware. This paradigm plays a pivotal role to achieve management flexibility of 5G services and reduce infrastructures’ investment and operational costs. Recent advances in artificial intelligence (AI) and the large amounts of data collected on orchestration platforms offer new perspectives. Many recent publications advocate the use of AI in the orchestration of virtualized networks and many algorithms are proposed. Though, today small amount of literature works implement or test these concepts in a real platform that considers network service chains data. In this work, we present a framework for Service Function Chains’ (SFCs) profiling and management through Machine- learning approaches. We propose an extended network orchestration platform based on Openstack, Kubernetes (K8s), and Open Source Mano (OSM) orchestrator. The benefits of this architecture are shown by an algorithm that realizes proactive auto-scaling procedure for service chains. It considers features’ importance per service type while achieving a trade-off between services’ stringent QoS requirements and the cost of resources usage.