Machine Learning Based SINR Prediction in Private Campus Networks

Mobile communication systems are in the deployment phase of their fifth generation (5G) and research activities toward the sixth generation (6G) of standards are ongoing around the globe. Integration of artificial intelligence and machine learning for the optimization of the network (e.g, radio access network functionalities) is one of the key technological enablers in modern mobile networks. Further, private campus networks (PCNs) are the local networks customised to serve certain vertical industry use-cases (e.g., smart warehouse and logistics). These PCNs are the key to the vertical market penetration of 5G standards and their proliferation. Optimizing the functioning of such PCNs is of key importance for both network operators and enterprises owning the PCNs. This paper presents machine learning based methods to predict signal-to-interference-plus-noise ratio (SINR) in the controlled environment of a PCN. The network measurements are collected from a live 5G SA PCN and a grid-wise radio environment map is generated. Subsequently, machine learning methods namely Vanilla long short-term memory (LSTM), Bi-directional long short-term memory (Bi-LSTM), and Random Forest are used in tandem with these measurements to predict SINR in advance based on mobility attributes of the user. The results show fair performance and demonstrate the applicability of such methods in PCNs.