Network State Awareness and Proactive Anomaly Detection in Self-Organizing Networks

Inference of network state and detection of anomaly network behavior based on the available data play important roles in the big data empowered self-organizing networks for enabling 5G. In this paper, we propose a novel framework of efficient network monitoring and proactive cell anomaly detection based on dimension reduction and fuzzy classification techniques. The enhanced semi-supervised classification algorithm allows adaptation of new behavior patterns, while incorporating a priori knowledge. The experimental results suggest that (i) our proposed method proactively detects the network anomalies associated with various fault classes, and (ii) the trajectory of the network states moving toward or away from a safe or fault class can be visualized, using the projected data onto a low-dimensional subspace.