• Clustering
• Time-series
• Anomaly Detection
• NFV
• Forecasting

Network Function Virtualization (NFV) is a cutting-edge paradigm in the telecom-
munications sector that enables network operators to offer adaptable and effective
services by adopting cloud computing principles into their infrastructure. In this
work, I developed a solution architecture that uses unsupervised learning together
with deep learning algorithms to forecast operation metrics and anomaly detection
for NFV. Our strategy entails first developing a cluster-based predictive model for
operation metrics forecasting and using it for anomaly detections. The solution pro-
vides network operations teams insight to support prompt proactive assessment and
decision-making. The analysis focuses on metrics at the infrastructure and service
levels. The former can be acquired directly from the monitoring system of an NFV
infrastructure, whereas the latter is commonly provided by the monitoring compo-
nents of the individual virtualized network functions. The solution architecture is
experimented using real-time data coming from the Vodafone NFV data center.