Cell Throughput Prediction Using AI Models: Insights from the O-RAN Framework

A massive number of connected devices in 5G/B5G leads to an ultra-dense network (UDN), causing several han- dover (HO) management issues. One of the critical functions in wireless network is load balancing which is distribute the traffic throughput equally across all Base Stations (BS). We will shed light on the Open Radio Access Network (O-RAN) fundamentals and architecture. O-RAN is a new era of mobile networks that uses open and interoperable Radio Access Network (RAN) nodes. The main objectives of the concurrent study are to engage artificial intelligence (AI) as an innovative solution for cell throughput forecast to detect load situation in advance to prevent high overload levels that can increase HO failure during the handover process. The paper studies the prediction of cell throughput using two AI models: i) a supervised deep learning (DL) model, which is Long Short-Term Memory (LSTM), and ii) a supervised machine learning (ML) model called XGBoost. The paper shows the difference in evaluation using Mean Squared Error (MSE) and training time by implementing state-of-the-art AI models as well as it compares the training time for both AI models. The study gives the first step to enhance handover management solutions.