A first view on mobile video popularity as time series
Aining Wang, Chen Zhang, Yuedong Xu · 2016
With the rise of mobile video streaming service, video popularity has drawn a great deal of attention in both academia and streaming industry. In this paper, we present a novel view of video popularity as time series. We collect more than two billion view records from one of the largest mobile streaming content providers in China. Using Discrete Fourier transform, we decompose popularity series of videos into two parts, the seasonal and residual component series. We find that the traffic dynamics of videos is governed by a small number of components in the frequency spectrum, and that traffic patterns within and across days differ from type to type. We further investigate the short-term and long-term stability of video popularity, and find that the stability metrics are highly correlated with total video requests.