Accurate user experience estimation through real-time throughput prediction with machine learning
Ahmed H. Eldeeb, Mohamed Nagah, Hesham Kamel, Wafaa Radi, Sara Fouad · Ain Shams Engineering Journal · 2025
This paper focuses on accurate throughput prediction for mobile broadband traffic to help Mobile Network Operators (MNOs) maximize network performance and to provide accurate estimation for user experience, Quality of Experience (QoE). In this work, we used real-world measurements from TEMS Investigation. We used drive test data to train different machine-learning models that can estimate user throughput across LTE network radio conditions. Our approach includes data collection, feature selection, model creation, and evaluation. A comparative study of thirteen machine learning methods predicts LTE data rates based on radio parameters, utilizing real-time throughput measurements for improved accuracy. The gradient boost with the hyperparameter tuning model has an R 2 accuracy of 87.8%, an RMSE of 1051.6 kbps, and an MAE of 688 kbps, proving our approach works. These findings show that our methodology can help MNOs optimize network resources, improve customer satisfaction, and reduce expenses.