Aplicações de handover inteligente em redes celulares baseadas em aprendizado de máquina

Abstract

This work aims to study, implement and evaluate different strategies to integrate Artificial Intelligence and Machine Learning models to handover procedures in cellular networks. The architecture proposals for the implementation are developed based on data from simulations of LTE networks with ns-3 simulator, and data collection campaigns on live network from a mobile network operator in Belo Horizonte. The observed results demonstrate great capacity of the models to produce outputs with high level of accuracy and diminished values of standard deviation, in both classification and regression proposed tasks. A wide range of models from the literature are encompassed in the evaluation, since more traditional ones, such as support vector machines, those based on tree architecture, artificial neural networks, as well as fuzzy systems, LSTM neural networks and latter boosting machines.