부산대학교 도서관

5G spectral efficiency requirements foresee net-work densification as a potential solution to improve capac-ity and throughput to target next-generation wireless net-works (NGWNs). This is achieved by shrinking the footprint ofbase stations (BSs), effective frequency reuse, and dynamic useof shared resources between users. However, such a deploymentresults in unnecessary handovers (HOs) due to the cell sizedecrements, and limited sojourn time on a high train mobility. In particular, when a train speedily passes through the BS radiocoverage footprints, frequent HO rate may result in seriouscommunication interruption impacting quality of service (QoS). This paper proposes a novel context-aware HO skipping thatrelies on passenger mobility, trains trajectory, travelling timeand frequency, network load and signal to interference and noiseratio (SINR) data. We have modelled passenger traffic flows incardinal directions i.e, north, east, west, and south (NEWS),in a novel framework that employs realistic Poisson pointprocess (PPP) for real-time mobility patterns to support mo-bile networks. Spatio-temporal simulations leveraging NEWSmobility prediction model with machine learning (ML) wheresupport vector machine (SVM) shows an accuracy of 94.51%. ML-driven mobility prediction results integrate into our proposedscheme that shows comparable coverage probability, and averagethroughput to the no skipping case, while significantly reducingHO costs.