Forecast analysis of call setup success rate in mobile networks: A hybrid approach

This study employs a hybrid forecasting approach, combining a statistical model and a machine learning model, to predict the monthly Call Setup Success Rate (CSSR) for Nigeria’s four major cellular networks using the CSSR dataset sourced from the Nigerian Communications Commission. The mobile networks studied were MTN, Airtel, Glo, and 9mobile. The machine learning model applied in this study was the Random Forest algorithm, while the statistical model used is the Autoregressive Integrated Moving Average (ARIMA) model. Model performance was evaluated using Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE), with forecasts validated against actual 2024 data. Results reveal consistently high CSSR values for most networks, with all operators frequently exceeding the Nigerian Communications Commission (NCC) benchmark. In this study, Random Forest generally outperformed ARIMA for datasets with pronounced non-linear fluctuations, while ARIMA performed better where trends were smooth and stable. Both models achieved exceptionally low forecast errors, with MAPE values below 0.5% for Airtel, Glo, and MTN, and below 3% for 9mobile using Random Forest. The findings demonstrate the viability of combining statistical and machine learning models for accurate KPI forecasting, supporting proactive network optimisation, regulatory compliance, and customer satisfaction strategies.