Improved Cellular Network Fault Prediction Using Machine Learning: In The Case Of Ethio Telecom Node B Network, Ethiopia

Abstract

In today's telecommunications settings, operators must deal with rapid technological developments while improving operational efficiency, which is, lowering operational costs while maximizing network performance. The growing amount of services available and subscribers has put cellular network service providers under a lot of strain. Subscribers to cellular networks have varying wants and requirements. Ethio telecom is one of the biggest telecom service providers in Africa and also the only one in Ethiopia. Its vision is to become a world-class telecom service provider. The primary aim of the research in this study is to relook at the concepts of cellular Network Fault Management proactive measure from a new perspective. The fault happened daily and need to rectify, so fault management regarding the faults of the cellular network needs proper attention and urgent solution. Because the company is currently using traditional methods of Network Management System (NMS), but in this thesis from drive test data, to improve ensure customer satisfaction, Cellular network fault of cellular must be managed properly. A Viable Solution to enhance its service available proactively is needed. It is, thus, with this intent that the researcher is motivated and decided to figure on the problem during this thesis. So far, most studies on Cellular network fault have focused on existing measurement of data available. Fault predicts were made by using real Ethio telecom EMS (Element Management System) data and achieved Previously done fault predicted using Time series considering down site NN(Neural Network) based and root cause analysis, however, the study had gaps, so the researcher was motivated to fill the gap. In this study, the researcher has tried to focus on the issue rather than existing measurement data from EMS (Element Management System) based to focus on No direct measurement data available. So improved prediction of Node B network fault empowered drive test. The result from DT (Drive test) data collected for two months shows that the performance of the Naïve Bayes with an accuracy of 98%