Machine Learning Based QoE Estimation Model for Video Streaming over UMTS Network
| dc.contributor.advisor | Yihenew, Wondie (PhD) | |
| dc.contributor.author | Digis, Weldu | |
| dc.date.accessioned | 2020-03-09T06:01:16Z | |
| dc.date.accessioned | 2023-11-04T15:13:11Z | |
| dc.date.available | 2020-03-09T06:01:16Z | |
| dc.date.available | 2023-11-04T15:13:11Z | |
| dc.date.issued | 2020-02 | |
| dc.description.abstract | The advent of data-intensive services needs quality Internet services. This in turn, makes Quality of Experience (QoE) gain prominent recognition in the telecommunications industry. Ethio telecom uses network Quality of Service (QoS) monitoring data obtained from Network Management Systems (NMS) tools to comprehend its network performances. However, as QoS measurement refers to network performances, this method does not generally give QoE data as perceived by the user. Therefore, QoE estimation models are proposed as solutions in the literature, recently. This study focuses on developing QoE estimation models using QoS features of round-trip time (RTT), jitter, loss rate (LR) and throughput, and QoE scores collected using Application for prediCting QUality of experience at Interne Access (ACQUA)-based crowdsourcing in Universal Mobile Telecommunication Systems (UMTS) networks in a real-time basis. Data preparations techniques such as data cleaning and dataset imbalance corrections have been applied to the collected datasets. Machine Learning (ML) algorithms of Arti cial Neural Network (ANN), KNearest Neighbor (KNN) and Random Forest (RF) are selected based on their suitability for multilabel problems. After training these models developed, they are evaluated using commonly used performance metrics such as accuracy, Root Mean Square Error (RMSE) and Receiver Operating Characteristics (ROC). Experimentation results exhibit that RF with an accuracy of 98.39%, is the best model while KNN and ANN achieve 87.47% and 77.59% overall accuracy, respectively. As a conclusion, all three models achieve acceptable performances. As a conclusion, our QoE estimation models if implemented can help Telecommunications Service Providers (TSP) in estimating user QoE in real-time. | en_US |
| dc.identifier.uri | http://etd.aau.edu.et/handle/123456789/21045 | |
| dc.language.iso | en_US | en_US |
| dc.publisher | Addis Ababa University | en_US |
| dc.subject | Universal Mobile Telecommunication Systems | en_US |
| dc.subject | Quality of Service | en_US |
| dc.subject | Quality of Experience | en_US |
| dc.subject | Supervised Machine Learning | en_US |
| dc.subject | Quality of Experience Estimation Models | en_US |
| dc.title | Machine Learning Based QoE Estimation Model for Video Streaming over UMTS Network | en_US |
| dc.type | Thesis | en_US |