PortalEyes: A Portal Identification Method via Noisy Video Traffic
Wei Min Yang, Zhenni Liu, Li Zhao, Zhongyi Zhang · 2022
As a new mode and new concept of modern city operation and governance, smart city has long-term significance for optimizing urban resource scheduling, improving urban operating efficiency, and improving the quality of life of citizens. As more and more citizens rely on video platforms(such as YouTube) for entertainment and information acquisition, media construction centered on video platforms has become an indispensable and important part of smart city, so the improvement of video service quality is becoming increasingly important. However, in the process of watching the video, the video portal and the video resource are transmitted separately, and it is difficult for the network providers to infer the portal (that is, the user's access intention) through the video resource. This paper abstracts the problem into the association of video resources and portals, and proposes PortalEyes, a portal identification method that combines co-occurrence relationship mining and order embedding, which quickly and roughly screens a large amount of traffic generated by users then sorts the screened results to further improve the accuracy. We conducted extensive experiments and compared the baseline and classic sorting algorithms. Our method makes the average position of the portal to be 1.2, which is the best compared with these algorithms. The experimental results show that our method is helpful for user intent discovery. It is of great significance to help Internet providers improve the quality of video services and improve the efficiency of information acquisition in smart cities.