Application of Big Data in Next-Generation Mobile Networks

This master thesis is dedicated to the research of different machine learning algorithms that can be used in LTE networks in order to achieve the best quality of service on each base station by influencing client handover procedure when the load is about to exceed the sufficient value. The thesis contains 69 pages and is based on 30 literature sources. The thesis includes 53 figures and 10 tables used to provide a better understanding of the topic and research results. The subject of the research consists of four time-series prediction algorithms: linear regression, XGBoost, Holt-Winters and SARIMA applied to forecast 4 consequent values of 3 self-organizing LTE network KPIs: RRC connections, PRB utilization and PCCH Utilization. These KPIs are forecasted in order to make the self-organizing network controller able to initiate the client handover in a predictive manner before the overload of the base station. This is different from the traditional approach where the controller acts reactively and applies configuration changes with up to 60-minute delay.