Analysis of the Correlation between User Behavior and User Engagement of Internet Video at Large-Scale
As the development of the video over the Internet becomes rapidly and the user scale becomes continuously extend, user expectations and requirements for high quality of service are constantly increasing, which makes service providers focus on the quality of user experience (QoE). And how to assess QoE accurately and effectively become the first problem to solve. Considering that there are many factors affect QoE, this paper studies the correlation between user behavior and QoE. In this paper, by use the method of statistical analysis and data mining, we make a research on user logs of large scale in Internet video based on a data set captured from one of the largest video operators of China. We measure behavior metrics such as forward, replay, pause, full screen, and we use user engagement to quantify QoE at a per view level. The study shows that user behavior can reflect QoE, especially the number of fast forward and replay are significantly associated with QoE. Finally, this paper uses k-means clustering algorithm to divide users into three groups with different behavior patterns according to user behavior, which indicator that user with different behavior patterns have different engagement. And we also find that when the session include more than 10 times of fast-forward, the engagement in the session will decline.
user behavior user engagement k-means big data Internet video
Yawei He Wenhui Zhang Weili si Anming Wei
Communication University of China, Beijing, China Academy of Broadcasting Planning, State Administration of Press, Publication, Radio, Film and Televi
国际会议
重庆
英文
186-192
2015-12-18（万方平台首次上网日期，不代表论文的发表时间）