A novel approach to predict latest trending videos from OVS using MLK (Multivariate Linear Regression)

Abstract

Predicting the top-N popular videos and their future views for a large batch of newly uploaded videos is of great 
commercial value to online video services (OVSs). Although many attempts have been made on video popularity 
prediction, the existing models has a much lower performance in predicting the top-N popular videos than that of 
the entire video set. The reason for this phenomenon is that most videos in an OVS system are unpopular, so 
models preferentially learn the popularity trends of unpopular videos to improve their performance on the entire 
video set. However, in most cases, it is critical to predict the performance on the top-N popular videos which is 
the focus of this study. The challenge for the task is as follows. First, popular and unpopular videos may have 
similar early view patterns. Second, prediction models that are overly dependent on early view patterns limit the 
effects of otherfeatures. To address these challenges, we propose a novel multifactor differential influence (MFDI) 
prediction model based on multivariate linear regression (MLR). The model is designed to improve the discovery of 
popular videos and their popularity trends are learnt by enhancing the discriminative power of early patterns for 
different popularity trends and by optimizing the utilization of multi-source data. We evaluate the proposed model 
using real-world YouTube data, and extensive experiments have demonstrated the effectiveness of our model.