Efficient Handover Algorithm in 5G Networks using Deep Learning

In 5G networks, microcells are densely deployed for the spatial reuse to cooperate with the traditional macrocells, and thus a moving user equipment (UE) usually experiences a more irregular change of the signal-to-interference-plus-noise ratio (SINK) and is more likely to disconnect to the serving cells when proceeding a handover. Hence, efficient handover algorithms in 4G networks no longer perform well in 5G networks. In addition, deep learning is a common method able to deliver highly accurate classification results for classification problems. In this paper, we make the first attempt to consider the SINK change of a UE in the handover problem in 5G networks, and to reduce the handover problem to a classification problem and solve the classification problem using a deep neural network (DNN). Simulations show that the proposed algorithm has a good performance in terms of the radio link failure rate and the ping-pong rate in 5G networks, as compared with the state-of-the-art methods.