Smart network monitoring system using AI-powered anomaly detection and performance optimization

Abstract

The aim of this project was to design and develop a Smart Network Monitoring System using AI-powered anomaly detection and performance optimisation to improve the efficiency, reliability, and security of computer networks. The project addressed the limitations of traditional network monitoring approaches, which are largely reactive, manual, and unable to proactively detect anomalies, leading to delayed fault detection, network downtime, and inefficient management. The system was developed using a mixed-methods design science approach involving requirements analysis, system design, implementation, and experimental evaluation. Network telemetry was collected through SNMP, Syslog, and NetFlow protocols and processed using an Isolation Forest machine learning model implemented in Python with scikit-learn, with results stored in MySQL and visualised through a PHP and Chart.js web dashboard. Evaluation against six simulated anomaly scenarios demonstrated a 100% detection rate, an overall F1 score of 0.93, a reduction in false positive rate from 38% to 2.9%, and an average mean time to detect of 25 seconds compared to over four minutes for the threshold-based baseline. The system successfully demonstrated that AI-powered monitoring is practically deployable in resource-constrained organisational environments and delivers measurable, operationally significant improvements over conventional monitoring approaches.