Back to work

STEG · Case 02 · 2025

Network Forecasting & Anomaly Detection

Can short, noisy hourly network histories produce forecasts and anomaly signals that are useful enough for operational monitoring?

Network racks and illuminated cables representing infrastructure monitoring
Role · Technical ownerSanitized case studyUtilities · Time-series modeling

The project

The challenge

I built a forecasting workflow and Streamlit application for hourly network traffic, covering robust ingestion, time-aware validation, classical and machine-learning comparisons, and anomaly overlays.

My role

I owned ingestion, exploratory analysis, feature construction, model comparison, temporal evaluation, anomaly logic, and the decision-facing Streamlit experience.

4.36Mbps RMSE
10usable series
6.2%RMSE reduction

Approach

Four clear steps

  1. 01Import
  2. 02Quality checks
  3. 03Backtest
  4. 04Monitor

Result

What the evaluation showed

Tuning lowered the evaluated hybrid XGBoost RMSE by 6.2%.

Initial4.65
Tuned4.36

Outcome

What changed

The result was a modular analysis application where ingestion quality, forecasting evidence, and anomaly context are visible in the same workflow.

Caveat

What this does not prove

Most evaluated histories covered only roughly twelve days of hourly observations.
Next projectCredit Card Default Prediction