Revisiting Popularity Characterization and Modeling of User-Generated Videos

Abstract

Abstract-This paper presents new results on characterization and modeling of user-generated video popularity evolution, based on a recent complementary data collection for videos that were previously the subject of an eight month data collection campaign during 2008/09. In particular, during 2011, we collected two contiguous months of weekly view counts for videos in two separate 2008/09 datasets, namely the "recently-uploaded" and the "keyword-search" datasets. These datasets contain statistics for videos that were uploaded within 7 days of the start of data collection in 2008 and videos that were discovered using a keyword search algorithm in 2008, respectively. Our analysis shows that the average weekly view count for the recentlyuploaded videos had not decreased by the time of the second measurement period, in comparison to the middle and later portions of the first measurement period. The new data is used to evaluate the accuracy of a previously proposed model for synthetic view count generation for time periods that are substantially longer than previously considered. We find that the model yielded distributions of total (lifetime) video view counts that match the empirical distributions; however, significant differences between the model and empirical data were observed with respect to other metrics. These differences appear to arise because of particular popularity characteristics that change over time rather than being week-invariant as assumed in the model.

Ephemeral Content Popularity at the Edge and Implications for On-Demand Caching

“…Although this second data collection took place well after the longitudinal collection was terminated, the crawls still provided information about 1,456,230 (60.86%) of the observed videos. Similar numbers have been observed by Islam et al [23], who were able to observe 67.13% and 55.23% of random (recently uploaded) and popular (keyword search) videos, respectively, when revisiting the YouTube site looking for videos from a prior data collection that had been carried out more than two years earlier [6]. Figure 1 shows the weekly viewing pattern over the duration of our longitudinal data collection period.…”

On the Dynamics of Social Media Popularity

“…Focusing on YouTube videos, Borghol et al [2011] showed how weekly based views can be used to model video popularity, and designed a model to determine the number of videos that may exceed some popularity thresholds. This work was recently revisited by Islam et al [2013], who showed that the weekly based modeling of popularity is still valid even years after video upload, but the synthetic model proposed for predicting the distribution of popularity of a group of videos is not. Zhou et al [2011] showed the importance of links to related videos to video popularity.…”

A survey into performance and energy efficiency in HPC, cloud and big data environments

“…It is noteworthy that other minor restrictions, not described here, were included in the query criteria in order to exclude irrelevant registers from the search results, such as conference presentations, journal editorials, and duplicated results. Additionally, it is important to mention that research papers similar to those of Birke et al 2013, Islam et al (2013) and Benmoussa et al (2013) were not part of this survey, although they regard systems' characterisations. This is due to the absence of the term 'workload characterisation' in their titles, keywords and abstracts.…”