Accurate prediction of downlink throughput in mobile cellular networks is critical for improving the quality of experience (QoE) in adaptive video streaming applications. With the growing demand for high-quality video content and the dynamic nature of wireless networks, predicting cellular throughput remains a challenging problem due to factors such as device mobility, diverse traffic patterns, and link variability. This study investigates the application of machine learning algorithms for throughput prediction, leveraging traffic traces collected from operational networks. We explore various algorithms, including Random Forest Regressor, XGBoost Regressor, Long Short-Term Memory (LSTM), a hybrid model combining multiple techniques, and ExtraTrees Regressor. The models are evaluated based on their predictive accuracy and ability to adapt to real-world network conditions, demonstrating the potential of AI-driven approaches for enhancing adaptive streaming systems. This work provides valuable insights into the practical deployment of machine learning for network performance optimization.

Keywords:

Cellular Throughput Prediction, Mobile Networks, Adaptive Video Streaming, Machine Learning, Random Forest Regressor, XGBoost Regressor, Long Short-Term Memory (LSTM), Hybrid Model, ExtraTrees Regressor, Quality of Experience (QoE), Traffic Traces and AI-driven Network Optimization.