Real-Time Network Monitoring, Performance and Traffic Analytics Platform: An Intelligent Framework for High-Frequency Telemetry, Statistical Anomaly Detection, and Operational Insights

Micro-bursts and temporary performance degradation can cause critical network bottlenecks before conventional monitoring systems detect them. Traditional approaches such as full packet capture are resource-intensive, while SNMP polling often fails to capture short-lived network events because of its relatively low sampling frequency. This paper presents a lightweight host-based real-time network monitoring and traffic analytics platform designed for high-frequency telemetry and anomaly detection. The proposed system collects network interface statistics at one-second intervals and uses active ICMP probing to measure round-trip latency and packet loss. Network performance indicators including transmit and receive throughput, link utilization, packet rate, average packet size, latency, and packet loss are continuously calculated. For anomaly detection, the platform combines a dynamic rolling Z-score method with an Isolation Forest model to identify both single-metric deviations and multivariate anomalies. A Streamlit and Plotly WebGL dashboard provides live visualization of telemetry, performance indicators, and alerts without requiring page reloads. Experimental evaluation using controlled network degradation scenarios demonstrates that the proposed hybrid detection approach achieves a 97.2% F1-score, 97.8% precision, 96.6% recall, and a 2.1% false alarm rate, with an average detection time of 1.15 seconds. The results demonstrate that the proposed platform can provide high-frequency network visibility and efficient anomaly detection while maintaining low system overhead.