MANAGING 5G LTE ADVANCED NETWORKS THROUGH MACHINE LEARNING INTELLIGENCE BASED SYSTEM

Abstract

The development of 5G mobile communication systems is aiming at meeting the boosting needs of mobile Internet traffic in the next 10 years. Some key technologies have been put forward in the next generation mobile communication technology (5G) such as millimeter waves and Small Cell to meet the requirements. The development of small cell networks offers higher chances in increasing capacity while reducing power consumption, since the small cells are effective supplements to macro cells with less power consumption in each single small cell. The problem is that deploying dense small cells may not actually save the total power of the system. To solve this problem, the small cells could be switched off during the time when traffic load is not that high, and switched on when large number of users appear in the hot spot area. However, switching off the small cells may lead to lower chance of offloading and harm the throughput. The performances of the algorithm are also evaluated by building a simplified LTE model with one macro cell and several small cells in MATLAB. Besides, the main traffic data of the network comes from the simulation in MATLAB which is considered as more realistic. To show the details of power consumption changing, we take the simulation time as a whole day (24 hours) while achieving and accuracy of 97.82%. By using the K-Means clustering machine learning based algorithm, the system power consumption can be largely reduced during peak hours and the decrement of throughput is within an acceptable range with management. The deployment of around 110 small cells in a macro cell shows the best performance.