Mobile Network Coverage Prediction Based on Supervised Machine Learning Algorithms

The need for wider coverage and high-performance quality of mobile networks is critical due to the maturity Internet penetration in today’s society. One primary drivers this demand dramatic shift toward digitalization Covid-19 pandemic impact. Meanwhile, emergence 5G wireless standard increasingly complex actual operating environment make traditional prediction model less reliable. With recent advancements promising capabilities machine learning (ML), it seen as an alternative approaches ground (G2G) communication prediction. In study, various ML models have been tested evaluated develop ML-based received signal strength networks. However, challenge identify a practical that can fulfill computing speed criteria while still meeting accuracy. A total six categories models, namely Linear Regression (LR), Artificial Neural Network (ANN), Support Vector Machine (SVM), Trees (RT), Ensembles (ET), Gaussian Process (GPR) consists more than 20 types established algorithms/kernels paper best contender among them, terms Findings from evaluation showed GPR most accurate Reference Signal Received Power (RSRP) $RMSE$ notation="LaTeX">$R^{2}$ , followed by ET, RT, SVM, ANN LR. Nevertheless, training times are also important factors determining RSRP several real-world network planning applications. Finally, ET with Random Forest (RF) algorithm has selected highly recommended practically employed developing rigorous predictions multi-frequency bands multi-environment. developed capable being utilized analysis optimization.