Video Streaming Data Traffic Prediction by Using Long Short Term Memory (LSTM) Model: In the case of UMTS Network in Addis
| dc.contributor.advisor | Mesfin, Kifle (PhD) | |
| dc.contributor.author | Begameder, Tamene | |
| dc.date.accessioned | 2020-03-09T05:30:20Z | |
| dc.date.accessioned | 2023-11-04T15:13:10Z | |
| dc.date.available | 2020-03-09T05:30:20Z | |
| dc.date.available | 2023-11-04T15:13:10Z | |
| dc.date.issued | 2020-02 | |
| dc.description.abstract | Predictive analysis of mobile network traffic is fundamental for the next-generation cellular network. Proactively knowing user demand allows telecom systems to perform optimal resource allocation. Nowadays, telecom companies face a network congestion problem; this problem results in longer delays, drastic jitter, and excessive packet losses. As a result, the quality of service (QoS) of networks deteriorates, and the quality of experience (QoE) perceived by end-users will be unsatisfied. As a solution, different researchers used statistical and neural network models for the prediction of video streaming data traffic. However, these models did not incorporate self-similarity and long term dependence characteristics of the video streaming data traffic. So, this study aims to predict the video streaming data traffic by using the Deep Learning, Long Short Term Memory (LSTM), model which incorporates self-similarity and long term dependence.We have reviewed various kinds of literature, conference papers, journals, white papers, and books related to the prediction of video streaming data traffic to achieve the objective of this study. Ten months of data (from October 2018 to July 2019) of video streaming data traffic information from five Radio Network Controllers (RNCs) of the Universal Mobile Telecommunication System (UMTS) network in the city of Addis Ababa (A.A) is collected. Finally, this research work result indicates that the LSTM model has 57.8% of MAE improvement of forecasting error compared to the hybrid model, i.e., Seasonal Auto-Regression Integrated Moving Average (SARIMA) and Extreme Learning Machine (ELM) model, which has the second lower error. The overall results of this research work demonstrate that the LSTM model is an effective method for predicting video streaming traffic to reflect temporal patterns. Such accuracy is vital to provide a better dynamic resource allocation for video streaming traffic. | en_US |
| dc.identifier.uri | http://etd.aau.edu.et/handle/123456789/21037 | |
| dc.language.iso | en_US | en_US |
| dc.publisher | Addis Ababa University | en_US |
| dc.subject | deep learning | en_US |
| dc.subject | forecasting | en_US |
| dc.subject | self-similarity | en_US |
| dc.subject | long term dependency | en_US |
| dc.subject | LSTM model | en_US |
| dc.subject | SARIMA model | en_US |
| dc.subject | ELM model | en_US |
| dc.title | Video Streaming Data Traffic Prediction by Using Long Short Term Memory (LSTM) Model: In the case of UMTS Network in Addis | en_US |
| dc.type | Thesis | en_US |