Classification of Network Slicing Requests Using Support Vector Machine

Abstract

As the 5th generation (5G) of wireless cellular communication systems are being rolled out around the world and with the rise of industry 4.0 verticals, different industrial sectors require different types of wireless network infrastructure. Network slicing concept provides a solution to this issue by allowing different architectures to logically co-exist on the same physical infrastructure. However, there is a need to automatically and correctly provision the network slice requests (NSR) based on their types. Many NS provisioning frameworks require the network slice requests (NSRs) to be pre-classified for optimal deployment. In this paper, a support vector machine (SVM) agent is proposed to classify NSRs into one of three network slice types: enhanced mobile broadband (eMBB), massive machine type communication (mMTC) and ultra-reliable and low latency communication (uRLLC). SVM has the ability to provide the most generalized classification on constructing an optimal separating hyperplane leveraging the nonlinear mapping of the input space into a higher dimensional space. The results show that accuracy has increase by up-to 4% when compared with other classification techniques.