Akademik Veri Yönetim Sistemi
2 nd International Karatekin Science and Technology Conference, Çankırı, Türkiye, 21 - 22 Aralık 2023, ss.4-9
In recent years, advancements in communication technologies have given rise to needs such as high transmission speed,
reliability, and low latency. Improvements in these aspects are crucial in fourth-generation (4G) communication technologies.
Following 4G, the Network Slicing method introduced with 5G allows the network infrastructure to be divided to meet
different service requirements, enabling flexible and efficient utilization of network resources. The performance of machine
learning-based 5G network slicing methods was tested by simulating 3rd Generation Partnership Project (3GPP) compliant
error-prone users and base stations. Five different machine learning methods, along with their parameter spaces, were used
in tests for network slicing, employing four methods (eMBB, M1oT, V2X, and URLLC). The performance of these classifier
models was analyzed using both error-prone user data and ideal user data. The simulation data were used to conduct a
performance analysis of machine learning methods mentioned in the literature, investigating their usability. A 96% accuracy
rate was achieved using the XGBoost method with error-prone user data, and a 97% accuracy rate was achieved with ideal
user data. Additionally, the relationships between the system cycle and user count, as well as the data rate reduction system,
were examined in the simulation