Section: New Results
Towards a Causal Analysis of Video QoE from Network and Application QoS
Participants: Michalis Katsarakis, Renata Teixeira, Maria Papadopouli (Univeristy of Crete), Vassilis Christophides
We have exploited an original framework for mining causal relationships among a 5-start rating of user QoE and various QoS metrics at network and application level. In particular, we have analysed QoE scores provided by a set of users for YouTube video streaming applications under different network conditions. We found that optimal QoE predictors we can be build using a minimal signature of only three features from application or network QoS metrics compared to four when features from both layers are considered. A thorough comparative analysis of the prediction accuracy of three models build using minimal signatures composed of (i) only network QoS, (ii) only application QoS, and (iii) both QoS features demonstrated that we can predict the QoE using only network QoS metrics and more surprisingly, predicting the QoE from network QoS metrics is as accurate as when using application QoS metrics. This work is the first step towards our ambition to assess QoE directly from network QoS metrics obtained via passive measurements of real traffic generated by online users. We will rely on the extracted minimal QoE/QoS signatures to build real-time predictors and compare their accuracy when using only network, only application or both QoS metrics. Last but not least, we plan to extend our experimental setting for other online applications such as teleconferencing services.