Performance Analysis And Network Optimization In LTE Based On S1-U Signaling Analysis

As the modern society steps into the era of Big data and mobile Internet, the volume of traffic data of the mobile terminal is soaring and it leads to the load of the mobile communication network increases greatly. Therefore, in order to provide better quality of service and user experience, the major operators are trying to construct LTE network. LTE network has been basically popular in the major cities in China now, on the basis of LTE network communication system is completed, maintenance and optimization is the most important work. LTE network is very complex, so we can not only rely on artificial operation deal with troubleshoot and repair to the communication network failure in time. The rapid development of artificial intelligence technology in this era provides the best way to let artificial intelligence replace manual operation in communication network maintenance and optimization. This thesis concentrates on the LTE network automation diagnosis and optimization work on the basis of this background.First, this paper analyzes the LTE network architecture deeply, each node of the LTE network is connected with the interface and the network protocol used in the interface. On the basis of this, the S1 -U interface and network protocol used in the S1-U interface are mainly analyzed. The development and related principles of machine learning in artificial intelligence technology and the application of Bayesian theory in various fields are briefly introduced. It is found that the Bayesian theory of machine learning can be combined with the LTE network to complete the automatic diagnosis system of the data plane of the LTE network.The principle of Bayesian and the Naive Bayes theory are analyzed emphatically and this paper verifies the performance of Naive Bayes Classifier by UCI repository. The common network faults in the LTE network are classified and the key performance indicator in the data plane of the LTE network are analyzed by the thesis. Finally, based on Naive Bayes theory, expert experiences and statistical data, network fault categories and key performance indicators, this thesis construct a simple Bayesian classifier finally. And the discretization of the key performance indicators with continuous variable attributes is completed by related discretization algorithm, so that all the key performance indicators are discrete variables. Finally, the fault automatic diagnosis system on the user plane of LTE network is constructed by the Naive Bayesian Classifier and the discretization algorithm.Through the fault automatic diagnosis system, we can get the root cause of the fault that there is overlapping coverage in the LTE network through the fault automatic diagnosis system. In order to solve the overlapping coverage problem,this thesis uses an antenna remote electrical tilt optimization algorithm based on the case-based Reasoning algorithm in machine learning in order to calculate the similarity between data and historical data. Finally, we can get the self-optimization algorithm of antenna Electrical tilt, the algorithm can reduce the coverage overlapping, improve the network throughput and optimize the user perception by the simulation results.