An ML-Informed Discrete Event Simulation Framework for Proactive Fault Management and Revenue Leakage Reduction in African Internet Service Providers

African internet service providers (ISPs) experience annual revenue losses of between USD 15 and 20 million due to poor fault management. Currently, 70% of faults are detected only after a customer complaint, and mean time to resolution is 8-12 hours-three times the global average. Although machine learning (ML) and Industry 4.0 methods are documented to reduce fault detection time by 30-50%, sub-Saharan African ISPs have not adopted these strategies due to data and financial constraints. A three-tier framework combining ML and discrete event simulation (DES) uses publicly available data from African telecommunications regulators and internationally established ML benchmarks to estimate revenue recovery under four proposed policies. The model, implemented in Python-based SimPy and run for 1,000 Monte Carlo iterations, evaluates three intervention policies for a Zimbabwean ISP against a Status Quo baseline. Results show that Policy A (revenue-prioritised fault dispatch) yields a 650% five-year return on investment for a USD 200,000 software investment, with payback under 12 months and USD 1.5 million in annual revenue recovery. Policy C (integrated ML and dispatch) recovers USD 4.8 million in annual revenue at a 160% five-year ROI. All results are robust to ±20% parameter variation. The framework and three-phase implementation roadmap are transferable to other ISPs across Sub-Saharan Africa and beyond.