5G Network Planning Optimization Using Machine Learning

In this paper, we present a regression-based machine learning (ML) optimization process to minimize the required number of 5G base stations for good coverage (>-80dBm received power) in a large percentage of the urban area. Area coordinates, orientations and states result in a total of 46 variables. To collect the data needed for training the ML algorithm, a Design of Experiment (DoE) algorithm Modified Extensible Lattice Sequence (MELS) is chosen. Using the data generated by MELS DoE, FAST (Fit Automatically Selected by Training) is used as ML algorithm to generate the ML model. Optimum coverage of the area is achieved with five base stations with 93.8% coverage greater than -80dBm.