부산대학교 도서관

Recently, video streaming has been one of the most popular application among mobile users. As the number of users and video streaming application increases, service providers are faced with the hefty challenge of ensuring high quality of service (QoS) given the lofty traffic load on the base stations. In this paper, we propose Machine Learning-Network Selector (ML-NetSel), an automated ML-based approach for selecting the best base station in a LTE environment. Our approach performs video traffic offloading at the level of base stations in high traffic scenarios to meet the QoS requirements of the various applications. It uses a hybrid approach to collect QoS requirements for users' applications along with users' behavior to train the ML-based model in order to select the base stations that can best serve these applications. Simulation results show that ML-NetSel provides higher prediction accuracy, computed using the Mean Absolute Percentage Error (MAPE) and higher throughput compared to an existing solution. It also reduces the delay and packet loss ratio.