Date of Award
ABR; Machine Learning; Video Streaming
Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.
While the internet video is gaining increasing popularity and soaring to dominate the network traffic, extensive study is being carried out on how to achieve higher Quality of Experience (QoE) in its content delivery. Associated with the HTTP chunk-based streaming protocol, the Adaptive Bitrate (ABR) algorithms have recently emerged to cope with the diverse and fluctuating network conditions by dynamically adjusting bitrates for future chunks. This inevitably involves predicting the future throughput of a video session. Predicting parameters being part of the ABR design, we propose to follow the data-driven approach to learn the best setting of these parameters from the study of the backlogged throughput traces of previous video sessions. To further improved the quality of the prediction, we propose to follow the Decision Tree approach to properly classify the logged sessions according to those critical features that affect the network conditions, e.g. Internet Service Provider (ISP), geographical location etc. Given that the splitting criterion will have to be defined together with the selection among ABR parameter values, existing Decision Tree solutions cannot be directly applied. In this thesis, some existing Decision Tree algorithm has been properly tailored to help learning the best parameter values and the performance of the ABRs with the learnt parameter values is evaluated in comparison with the existing results in the literature. The experiment shows that this approach can improve the performance of an ABR algorithm by up to 8.59%, with 98.38% of the testing sessions performing better than having a fixed parameter value, and only 0.8% performing worse.
Le, Anh Minh, "Improving Adaptive Video Streaming through Machine Learning" (2018). Electronic Theses and Dissertations. 7373.