Data Mining Approach to Analyze Mobile Telecommunications Network Quality of Service: The Case of Ethio-Telecom
| dc.contributor.advisor | Jemaneh, Getachew (PhD) | |
| dc.contributor.author | Deyu, Lulu | |
| dc.date.accessioned | 2018-11-27T13:11:39Z | |
| dc.date.accessioned | 2023-11-29T04:56:52Z | |
| dc.date.available | 2018-11-27T13:11:39Z | |
| dc.date.available | 2023-11-29T04:56:52Z | |
| dc.date.issued | 2014-05 | |
| dc.description.abstract | Huge amount of measurement data indicating the performance of a mobile network has been generated. Sometimes it is very difficult to draw essential information from this complex data merely applying domain expertise and prior knowledge. For the ultimate goal of QoS improvement, it is helpful to follow a data mining approach to deal with this complex data. In this study, a sample data on three time stamps such as 24-hour, 1-month, and 1-week day and night high traffic hours, indicating QoS KPIs has been taken from the live network of ethio telecom. Strictly following the KDD process, various experiments are conducted using the Weka open source data mining tool. This is done to find out the number of clusters that logically segment the KPI data applying the simple k-means algorithm and the best classification model comparing the J48 decision tree, the Naïve Bayes, as well as the Multilayer Perception classifiers. It has been found that the cluster model worth splits the KPI data in to five clusters on the basis of their natural proximity. These clusters are ranked and labeled to be applied on the next classification model experiment. In this experiment a data set of four selected attributes and 8478 instances has been used to build and select the best model. A separate data set of 4240 instances is provided to finally evaluate the classification accuracy of the selected model for unseen data. As a result a classification model built on Multilayer Perception with 6 ‗Hidden Layers‘, ‗Learning Rate‘ of 0.1 and ‗ Seed‘ value of 2 has got the best classification accuracy by correctly classifying 84.4953 % of the data in to their classes. Key Words: QoS, KPI, KDD process | en_US |
| dc.identifier.uri | http://etd.aau.edu.et/handle/123456789/14561 | |
| dc.language.iso | en | en_US |
| dc.publisher | Addis Ababa University | en_US |
| dc.subject | QoS | en_US |
| dc.subject | KPI | en_US |
| dc.subject | KDD process | en_US |
| dc.title | Data Mining Approach to Analyze Mobile Telecommunications Network Quality of Service: The Case of Ethio-Telecom | en_US |
| dc.type | Thesis | en_US |