Deep Learning-based Reference Signal Received Power Prediction for LTE Communication System

A highly accurate prediction of radio signal power is crucial for planning the coverage of mobile networks. Currently, a path loss model is most widely used to predict the radio signal. However, the path loss models commonly provide an over- or under-estimation of the signal power. In this paper, we present the reference signal received power (RSRP) prediction using a deep learning. To evaluate the performance of our prediction system, we use the empirical data in Bangkok metropolitan area. Especially, the empirical data comprise 2 million measurements per day for deep learning. The root mean square error (RMSE) value of our prediction is approximately 3.91 dB.