Unified Real-time Framework for Proactive Network Health management

In an era of fast-growing reliance on interconnected network infrastructures, ensuring seamless performance and reliability has become paramount. The scenario demands more secure and real-time networking health monitoring to provide quality of service (QoS). Existing network health monitoring systems often lack integrated capabilities for both detecting current anomalies and predicting future issues.This paper introduces a unified framework using machine learning for real-time network health classification and traffic anomaly detection. The approach makes use of data preprocessing, including Principal Component Analysis (PCA) for dimensionality reduction, followed by classification using K-Nearest Neighbors (KNN), Random Forest (RF), and Support Vector Machine (SVM) models. The framework was evaluated using the CICIDS2017 dataset by a standard train/test split methodology. The findings show a high classification accuracy, with KNN achieving 99.46%, RF 96.68%, and SVM 88.18%, effectively distinguishing between benign and various types of attack traffic. The results showcase the potential of the proposed framework to enhance network resilience and operational efficiency, offering proactive management capabilities suitable for next-generation networks (NGNs).