A Machine Learning-based xAPP for 5G O-RAN to Mitigate Co-tier Interference and Improve QoE for Various Services in a HetNet Environment

Data traffic has skyrocketed as a result of the global proliferation of rich media services. A number of cutting-edge applications are predicted to be supported by 5G networks across the three categories: enhanced mobile broadband, ultra-reliable low latency communications, and enormous machine-type communications. The expectations and goals for the various services in the 5G networks have put a lot of pressure on mobile operators to maintain high Quality of Experience (QoE). The use of 5G Heterogeneous Networks (HetNets), which will provide consumers with the ability to be associated with either Macro Base Stations (MBS) or small cells, is one of the most promising solutions. Among the small cells, femtocells have drawn much attention recently. Yet, the most significant challenge with the deployment of femtocells is the high co-tier interference that can occur between different femtocell users. Artificial Intelligence (AI) and Machine Learning (ML)-based solutions are being incorporated in 5G networks to address this challenge. In this paper, we propose a ML Multi-Classification and Offloading Scheme (MLMCOS) to mitigate co-tier interference in 5G HetNets. MLMCOS classifies users into multiple classes based on their service priority along with their experienced co-tier interference. It then offloads some of them to the nearby Femto Base Stations (FBS) based on the availability of resources to ensure high QoE. ML classification algorithms are evaluated in terms of accuracy, recall, and precision. The performance of MLMCOS is then compared to those of Proportional Fair (PF) scheduling algorithm, Variable Radius and Proportional Fair scheduling (VR+PF) algorithm, and a Cognitive Approach (CA) in terms of Video Multimethod Assessment Fusion (VMAF), R-Factor, and RUM Speed Index (RUMSI).