Digital Twin of a Campus Network for Traffic Forecasting and Anomaly Detection

A university IT office asks three questions about its Internet uplink: how much traffic arrives tomorrow, is today's traffic normal, and when does the link need an upgrade? This project builds a small digital twin of real academic networks from the real hourly traffic of 40 institutions of the Czech research and education network CESNET (CESNET- TimeSeries24, Zenodo 13382427, CC BY 4.0). It forecasts every next day with classical methods (seasonal naive, SARIMA) compared with LSBoost, a global LightGBM, a small LSTM and the Chronos-Bolt tiny foundation model used zero- shot; it raises alarms with the classical 3-sigma rule, forecast-residual z-scores, CUSUM and an Isolation Forest, evaluated on injected (labelled) anomalies because the data have no incident labels; and it runs capacity what-ifs (traffic growth, link upgrade) with an M/M/1 queue on assumed link capacities. A web application on five databases shows the twin per link. A GNU Octave cross-check, a MATLAB version and an InfluxDB + Grafana dashboard are included.

Category: Information Science and Technology

Information Science & Technology Project IST03 from ProTechSC Academy, with graphical abstract, video demonstration and project guidance.