DataraFlow 2026
From manually collecting telecom data to refining my research topic, this week taught me that Data Science begins long before writing a line of code.
I thought I was beginning to understand Python.
Week 2 of my DataraFlow journey is officially in the books.
What I learned, built and struggled with during Week 1 of my self forged learning journey.
Mobile network congestion remains a persistent quality of service problem for Nigerian telecommunications operators, with regulatory data recording 413,587 consumer complaints in the first half of 2026 and a Q1 2026 industry capacity report showing MTN's urban network congestion rising from 4.0% to 6.7% between September 2025 and February 2026. This project investigates whether supervised machine learning can predict TCH Congestion, the Nigerian Communications Commission's own radio access network metric for traffic channel overload, at the state level for MTN, Nigeria's largest mobile operator. Using real monthly Radio Access Network KPI data published directly by NCC, covering all 37 Nigerian states across January 2025 to January 2026, this study develops and compares regression based models, including Random Forest and Gradient Boosting, against a naive historical average baseline, to forecast next month's TCH Congestion per state using prior congestion trends, state, and network generation as features. SDCCH Congestion, the related control channel metric capturing call setup failures before connection, is evaluated as a secondary target to test whether early warning signals improve prediction accuracy. Models are evaluated using RMSE, MAE, and R² against held out, chronologically split data. The expected contribution is a reproducible, state level forecasting pipeline using an operator identified, regulator published dataset, offering a practical basis for anticipating where MTN's network capacity is likely to come under strain before subscribers experience degraded service.