Predicting Channel Quality Indicator (CQI) in LTE Using Ensemble Learning Approaches

Correct prediction of Channel Quality Indicator (CQI) is a key to successful link adaptation and resource allocation in Long-Term Evolution (LTE) and 5G New Radio (NR) networks. The paper presents a CQI prediction method based on an ensemble-based technique by using measurements on the LTE radio signals like Reference Signal Received Power (RSRP), Reference Signal Received Quality (RSRQ), Received Signal Strength Indicator (RSSI), and Signal to Noise Ratio (SNR). The proposed method is evaluated against with individual machine-learning models since it combines complementary learners to improve prediction stability and accuracy. Result obtained from publicly available datasets of LTE show the proposed framework outperforms other competitors with an MAE of 0.66 and R2 of 0.93. The ensemble would add to training and inference time, but prediction latency per sample (0.052 ms/sample) is much lower than LTE timing requirements, and thus quite practical to use in real time. In general, this work demonstrates that an ensemble-based method can provide a significant boost to the efficacy of CQI estimation, which can become a promising solution to effective scheduling and adaptive modulation of LTE systems.