A Channel Quality Prediction Approach for a Commercial Multiband Mobile Nertwork

Accurately predicting wireless channel quality is crucial for enabling mobile network operators to carry out proactive network operations. In this article, we address the challenge of predicting channel quality across various wireless links. Building on our previous work, we introduce two new models: a Convolutional Neural Network (CNN) and a hybrid CNN-LSTM, which combines CNN with a Long Short-Term Memory (LSTM) architecture. These models are designed to predict the Channel Quality Indicator (CQI) in 4G LTE/5G networks incorporating small cell base station architectures. We evaluate their performance using a dataset collected from Orange Senegal’s commercial 4G LTE/5G network. Our results demonstrate that the CNN, LSTM, and CNN-LSTM models adapt effectively to real-world conditions and achieve high prediction accuracy. Among them, the CNN-LSTM model delivers the best performance (RMSE = 0.25), followed by the LSTM model (RMSE = 0.281), and the CNN model (RMSE = 0.308).