Understanding the Usage Characteristics of Personalized Online Video Services

Personalized online video services, including personalized live streaming and personalized short video services, have gained immense popularity and social influence. Understanding their usage characteristics is fundamental for service providers, app developers and network operators. Using a dataset of over 2 billion access logs from a cellular network, this paper examines the usage pattern and user behavior of personalized online video services. Our key observation is that, in spite of the significant differences in used devices and interests in other apps, the users in both personalized live streaming and personalized short video services are mainly entertainment oriented. Besides, these two services show limited association with the traditional online video service. We also reveal that the user moving range is a key factor that influences user engagement in personalized short videos, but not in personalized live streaming. Other observations include biased spatial usage distribution and heavy-tailed usage time distribution for both types of services.