@Article{cmc.2026.076246,
AUTHOR = {Muhammad Mukhtar, Farizah Yunus, Ahmad Shukri Mohd Noor, Zulfiqar Ali, Muhammad Junaid, Mehmood Ahmed},
TITLE = {Handoff Decision-Making in 5G Cellular Networks Using Deep Learning},
JOURNAL = {Computers, Materials \& Continua},
VOLUME = {87},
YEAR = {2026},
NUMBER = {3},
PAGES = {--},
URL = {http://www.techscience.com/cmc/v87n3/66921},
ISSN = {1546-2226},
ABSTRACT = {The increasing adoption of 5G cellular networks has introduced significant challenges for network operators. The main challenge lies in the management of seamless handoff (HO), which occurs owing to the rapid expansion of equipment, data, and network complexity. To address this challenge, a model named optimal HO management deep learning neural network (OHMDLNN) is proposed. The model is trained on network activity data, and it uses KPIs (key performance indicators) and system-level parameters to make HO decisions. As demonstrated in the article, OHMDLNN is successful in analyzing the effect and interdependence of KPIs from both the network and user equipment (UE) perspectives. Moreover, the model is evaluated for accuracy (the percentage of correct decisions made by a model on a dataset) in comparison with existing neural network-based HO decision models. These include temporal convolution networks (TCN), recurrent neural networks (RNN), long short-term memory (LSTM), gated recurrent units (GRU), and convolutional neural networks (CNN). The dataset used to evaluate the performance of the model consisted of 65,000 records. The model demonstrates superior performance, with an average improvement on accuracy of 8 percent over TCN, 18 percent over RNN, 6 percent over LSTM, 14 percent over GRU and 4 percent over CNN. Along with accuracy, the model is also tested on important performance indicators, including the packet loss rate, the success rate, latency, and throughput at the time of handover. These results affirm its efficiency in the HO decision-making process. Future research will consider the use of advanced deep learning architectures and simplify the process of integrating system-level inputs to optimize system performance during HO events.},
DOI = {10.32604/cmc.2026.076246}
}