A Two-Stage Clustering Method for P2PTV Traffic Classification

DOI

Bibliographic Information

Other Title
  • P2PTVトラヒック分類における2段階クラスタリング方式

Abstract

This paper proposes a two-stage clustering method to anlyze the traffic generated by peer-to-peer-based video delivery service (P2PTV), which has recently attracted attention. Since P2PTV video contents tend to be long, peers will join or leave dynamically, and thus the characteristics of P2PTV traffic will vary momentarily. The authors divide P2PTV traffic into data units and try to extract the unit traffic as components of P2PTV traffic by using clustering, a machine learning technique. However, clustering for all the data units generated by multiple video contents does not work appropriately due to the bias of occurrence rate. The proposed method performs the first-stage clustering on each video content, then creates the feature elements of the obtained clusters, and finally performs the second-stage clustering. Using 100 sets of P2PTV video traffic data, we also analyze and evaluate the classification characteristic of the proposed method comparing to the conventional one-stage clustering.

Journal

Keywords

Details 詳細情報について

Report a problem

Back to top