An Efficient Network Slice Allocation in 5G Network Based on Machine Learning

Abstract

The 5G network is a very flexible network that not only provides connectivity between mobile devices but also connects smart homes, smart industries, and autonomous vehicles. 5G offers diverse services like massive machine type communication, enhanced mobile broadband, and ultra-reliable and low latency communication. To deliver these services to end-users, network slicing plays a vital role. To offer different services to different users simultaneously and make the appropriate selection of network slices is a challenging task. In this paper, machine learning models with or without employing feature selection algorithms are used to predict the best network slice for each incoming request. Network functions virtualization and Software-defined networking are playing an important role in implementing network slicing in the 5G network, are also presented in this paper. The results of various classification models without employing feature selection and with the Analysis of Variance (ANOVA) feature selection algorithm are compared with some existing ML models, and the proposed model ANOVA+ANN got 94.46% accuracy, which is much higher than other existing models. The main goal of this research is to make network slicing work well in the 5G network so that it can offer many different services over a single physical network.