Characterizing and modelling popularity of user-generated videos

Abstract

This paper develops a framework for studying the popularity dynamics of user-generated videos, presents a characterization of the popularity dynamics, and proposes a model that captures the key properties of these dynamics. We illustrate the biases that may be introduced in the analysis for some choices of the sampling technique used for collecting data; however, sampling from recently-uploaded videos provides a dataset that is seemingly unbiased. Using a dataset that tracks the views to a sample of recently-uploaded YouTube videos over the first eight months of their lifetime, we study the popularity dynamics. We find that the relative popularities of the videos within our dataset are highly non-stationary, owing primarily to large differences in the required time since upload until peak popularity is finally achieved, and secondly to popularity oscillation. We propose a model that can accurately capture the popularity dynamics of collections of recently-uploaded videos as they age, including key measures such as hot set churn statistics, and the evolution of the viewing rate and total views distributions over time.

Popularity Growth Patterns of YouTube Videos - A Category-based Study

“…Although the most similar dataset collected (Borghol et al, 2011) shows that the views of Music category exceeds all other categories within their 8-month measurement period, 3 our dataset shows that popular News, and Sports videos enjoy higher viewing rates than any other types of videos for the first couple of days since publication. Figure 4 suggests that almost all categories have at least 5% of their videos that experience a high initial viewing rate; the difference is that after these few peak days, views for most of the categories become very low, except Music and to a lesser extent, Film and Tech videos.…”

“…Recent work was done on nearly 30,000 videos, collected by using the recently uploaded standard feed provided by the YouTube API (Borghol et al, 2011). Their collection procedure claims to have an unbiased dataset; the Most Recent standard feed returns video information randomly that are uploaded recently.…”

“…We modified previous unbiased data collection methods (Borghol et al, 2011) since we speculate that the first week since uploading deserves more investigation, even though this may expose day-of-week effects. Moreover, similar numbers of videos from all the categories are needed for appropriate comparison between different categories.…”

“…One consequence of this restriction is that the number of videos in each category is significantly reduced, down to 42% for News and Sports and 18% for Animals and Travel. We define time-to-peak as the day in which a video experienced the most views as in previous work (Borghol et al, 2011). Time to reach peak popularity is not the same for all categories.…”

“…Much research has been done investigating request characteristics from both client (Gill et al, 2007;Zink et al, 2009) and server perspectives (Borghol et al, 2011;Cha et al, 2009;Ding et al, 2011;Figueiredo et al, 2011) in order to enable improved service. However, none of this earlier work considered the types of video objects.…”