A short video recommendation algorithm incorporating network awareness and video bitrate

With the popularity of short video platforms, a large number of users have been attracted, bringing significant benefits to the platforms. However, traditional recommendation systems, due to their failure to consider network conditions, may cause a few key recommended videos on the user’s homepage to fail to be transmitted due to their high bitrates, thereby reducing the user’s viewing experience and the effectiveness of recommendations, and further leading to a decrease in the platform’s revenue. Therefore, this paper designs a NSRec short-video recommendation algorithm. The algorithm combines the video bitrate with the current network-aware transmission prediction module, integrates it with the short-video quality evaluation module, and adapts the fixed recommendation quantity based on the multi-objective rearrangement module according to different scenarios. Ultimately, the maximum benefit of video playback was achieved, and the core issue of the disconnection between the recommendation system and network transmission was resolved. Experiments based on Kuaishou dataset and multiple application scenarios show that NSRec has significant performance, with the average video playback duration of users increased by 20 to 160%, and the transmission success rate of recommended videos increased by 43%.