The New Neural Network Based Framework for Fading Channel Prediction for 5G

The paper presents a developed framework based on several neural networks with fine tuning for predicting CQI based on the time series of SINR values in order to select the best MCS on the eNB side, taking into account fading in the radio channel in downlink. A comparative analysis of the performance of modern architectures in solving the problem of forecasting time series and substantiating the choice of LSTM as the core of the predictive framework is performed. A comparative analysis of the learning algorithms of the architectures under consideration is performed. The choice of the model is justified and the performance evaluation of the developed framework is given. The choice of parameters of the developed predictive framework is justified.