ML-Based Self-Optimization Handover Technique for beyond 5G Mobile Network

The Fifth Generation (5G) and Beyond (B5G) mobile systems employ advanced technologies, such as millimeter Wave (mmWave) and Ultra-Dense Networks (UDNs), to meet future networks' requirements. However, implementing these technologies may pose several challenges to the B5G network. One key challenge is the need for efficient Handover (HO) optimization processes. HO aims to ensure seamless connectivity and uninterrupted services for users while moving from one cell to another within the coverage area. Thus, this study introduces a new, intelligent, and robust self-optimization HO technique designed to work efficiently with the B5G networks. The technique utilizes Machine Learning (ML), particularly leveraging the Regression Tree (RT) model. In this study, the proposed technique is referred to as the ML-based Self-Optimization Handover Technique (ML-SOHOT). The technique is evaluated and validated using different major HO metrics, including Handover Probability (HOP), Handover Failure (HOF), and Handover Ping-Pong (HOPP) across various mobility patterns in B5G, specifically considering an urban environment. The results demonstrate that ML-SOHOT enhanced the HO optimization performance significantly and surpassed the competitive algorithms. Furthermore, ML-SOHOT achieves an average HO performance improvement of up to 96% compared to competitive algorithms from the literature. Consequently, the technique could enhance the overall B5G system performance and user experience.