Improving resource allocation in 5G networks using traffic segmentation based on machine learning techniques

Due to a significant increase in cellular network traffic, predicting network traffic using traditional methods may lead to inaccurate allocation of available resources. Current and future cellular networks target ultra-low latency, high reliability standards, improved security, better capacity, as well as efficient user’s communi-cations. This work adopts 5g network slicing technology to respond to different users’ requirements. The optimization of resource allocation to network slices to meet different network traffic is of great demand. Therefore, this work focuses on the implementation of an algorithm of network slicing based on machine learning in order to group IoT devices in 5G networks into three efficient network catego-ries, namely eMBB, URLC, and mMTC, according to the traffic. We utilized KNN, SVN, and LR machine learning algorithms to classify devices according to use cas-es within the three aforementioned segments. Results show that these algorithms perform excellently in predicting the best suitable slice for the network traffic quality. The basic metrics of performance, including accuracy, F-score, and sensi-tivity are examined. Comparative analyses illustrate that KNN, SVN, GNB, and LR have the ability to classify network traffic slices with an accuracy of up to 95%.