Machine Learning–Based Prediction of Channel Stability for Efficient CQI 5G Networks

Abstract

Whether the 4G and the 5G mobile networks are effective or not is highly interconnected with the management of signaling overhead, especially Channel Quality Indicator (CQI) reporting. Regular updates in CQI, though necessary to ensure reliability, may lead to traffic congestion in the signaling which may impact on the overall network performance. This paper introduces a machine learning-driven scheme that minimizes cases of unneeded CQI transmissions by directly predicting channel stability based on CQI data. The analysis of prediction accuracy under different conditions is achieved by using modeling, e.g. Convolutional Neural Networks (CNNs) and Support Vector Machines (SVMs). The findings reveal that CNNs are always more effective in predicting channel stability than SVMs, which proves that they are the most appropriate in terms of optimizing adaptive CQI reporting. The proposed solution is not only able to guarantee the efficient signaling, but also the basis of the next-generation networks on low-latency and high-throughput communication.