Applying data mining techniques for predicting telecommunication service faults: the case of Ethio-telecom

No Thumbnail Available

Date

2013-10

Journal Title

Journal ISSN

Volume Title

Publisher

Addis Ababa University

Abstract

Faults are inevitable in telecommunication services, therefore predicting them ahead of time is crucial to make the systems more robust and the operation more reliable. Faults in telecommunication services have direct impact on its availability and maintenance costs, so their quick elimination, prevention and removal of causes that generated them, is of special interest. This study is aimed at applying data mining techniques to support prediction of broad band network service faults at Ethio-telecom. The subject of this study is from Ethio telecom's Z-Smart Trouble Ticket system, which contains customer's service fault report information and remarks given by experts about the actual fault reasons after the problems solved. In the data mining process, the first step was collecting the target data from the above mentioned system at Ethio-telecom. Then various types of preprocessing tasks were performed on the collected data so that to make the data ready for the planned data mining tasks. On the model building phase. (4.5 variant of decision tree and Naive byes of Bayesian network algorithms were applied for building the classifiers and accuracy results obtained using J48 and naïve byes was 74•06% and 69% respectively. Due to the data set imbalance observed on the class variables, SMOTE minority over sampling technique with J48 algorithm was applied and it improves the classifier accuracy to 77•90%. The results from this study were encouraging, which strengthened the belief that applying data mining techniques could in fact support network service faults prediction activity at Ethio telecom. In the future, using a balanced data set, and incorporating more attributes and also by testing with various classification algorithms better classifier accuracy could be obtained

Description

Keywords

Data Mining, Classification, Prediction, Telecommunication, Network, Faults.

Citation