Characterizing User Watching Behavior And Video Quality In Mobile Devices

With the development of mobile video, more and more online video content providers began offering client which can be installed on a smartphone or tablet, so that users can browse the site in any wireless network environment for watching videos. Mobile video playback systems give users great convenience, with the popularity of mobile video viewing and smart mobile terminal developed, mobile video traffic will became a major component of mobile network traffics. However, there are still unknown about how users watch video and which factor influence on video viewing quality in different mobile devices and different network access.Based on a large-scale dataset extracted from the servers’ logs of PPTV, one of the largest online video service providers in China, this paper studied how device types, wireless network connections and video qualities impact user’s watching behaviors and network traffic.The contribution of this paper are mainly in the following:(1) build Hadoop distribution system to analyzed user behavior in mobile devices, the founding are mainly about:the diurnal user viewing patterns on mobile devices are different from that on PC devices in peaks; the mobile APP’s category page is the primary source for users to find videos to watch, and the keyword search is the secondary source, we suggest that the category page contribute for video viewing diversity and the search engine is more purposeful; users using mobile devices watch shorter than users with PC devices, but are more concentrated on popular videos so that the video popularity distribution is more skewed, found that users’watch time distribution can be best fit in a power-law module;(2) analyzed video quality in different access type and mobile terminal, found that iOS devices are better than Android devices and tablet are better than smart phone;(3) after the observation and analysis of mobile video viewing quality, we found that the metrics of video viewing quality (video start latency, video buffer ratio and success ratio) are uncorrelated and their distributions are uniformity over time;(4) at each of the video viewing quality factors(ISP, geographic, devices, access type and video resolution), the video resolution and the geographic of users’are more important than other factors and the Internet services provider is less important than other factors. Thus we can use a minimum deployment by adjusting the video bitrate of video transmission system or arranging geographic location contents cache system to achieve a maximum benefit of video viewing quality. For the current user’s mobile video viewing behavior and the quality of services, we further provide insights and suggestions in providing mobile video services and improving services’quality.