Comparative Analysis of ARIMA, Prophet, and Glmnet for Long Term Evolution (LTE) Base Station Traffic Forecasting

This study evaluates the performance of three forecasting models—ARIMA, Prophet, and Glmnet—with the primary objective of equipping the telecommunication industry with effective tools for cellular traffic forecasting. These tools lay the foundation for efficient resource management, cost optimization, and enhanced service delivery. The study begins with dataset description and preparation, followed by the selection of traffic forecasting models, and concludes with performance evaluation based on metrics such as Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), Symmetric Mean Absolute Percentage Error (SMAPE), Root Mean Squared Error (RMSE), and Coefficient of Determination (R²). The main contribution of this research is a comprehensive comparison of the three forecasting methods, aiding practitioners and researchers in identifying the best prediction model for specific contexts. The findings reveal that Glmnet consistently outperforms ARIMA and Prophet across all categories of traffic forecasting on the selected performance metrics. Its ability to handle complex data structures, manage multicollinearity, and deliver robust and accurate predictions makes it the preferred choice for forecasting cellular network traffic in the telecommunications domain. Doi: 10.28991/ESJ-2024-08-06-04 Full Text: PDF