Drive Test Minimization Using Deep Learning with Bayesian Approximation

Drive testing is a common practice performed by operators to optimize and evaluate their mobile networks with respect to capacity and coverage. For dense areas, drive test measurements are very time-consuming due to many obstacles causing Non-Line-Of-Sight (NLoS) scenarios. In this paper, we show how Deep Learning (DL) techniques can be utilized to predict LTE signal quality metrics using drive test measurements. Moreover, we show how the obtained solution can offer insight into where additional measurements are required. The proposed solution can accurately predict LTE signal quality metrics reducing drive tests needed by up to 70%.