Asia-Pacific Network Operations and Management Symposium
P2PTV Traffic Classification and Its Characteristic Analysis Using Machine Learning
Koji Hayashi, Rina Ooka, Takumi Miyoshi, Taku Yamazaki,
PDF download (1.1MB)
This paper proposes a classification method for peer- to-peer video streaming (P2PTV) traffic using machine learning. Since the user terminals (peers) share video data in P2PTV, P2PTV traffic is difficult to control and manage statically as both the number of peers sharing the same video data and the throughput vary with respect to contents. Although there exists a conventional method to classify and model P2PTV traffic by focusing on the number of peers and throughput, problems on the classification criteria and reproducibility remain in this method. In this paper, we use a clustering method that is considered as one of the machine learning methods and try to classify P2PTV traffic data into some categories. We extracted 18 features by analyzing P2PTV traffic of popular P2PTV applications: PPStream and PPTV; and then classified the traffic. The classification results show that about 400 traffic data sets were categorized into four clusters.