Multidimensional linear regression analysis to predict the change in one-day video views on Bilibili website
Jixing Lu, Yifan Xia · Highlights in Science Engineering and Technology · 2024
This paper utilizes a multidimensional linear regression analysis model to investigate video playback volume changes within Bilibili's film, cartoon, anime, auto-tune remix-themed content, game, and entertainment partitions. We employ crawler software and a webpage loop with a nested collection of video links to extract and sequence titles. Technical abbreviations are defined when first presented. We utilize crawler software to extract various data sets from online video links. We establish a web page cycle and collected video links within nested collections to accomplish this. Subsequently, they systematically accessed each link to extract information such as titles, playback, likes, coins, favorites, and video synopses. Finally, the data 6,075 are obtained and filtered down to 1,691 representative data points using playback, likes, coins, favorites, and other data sets from the previous day. The previous day's playback, likes coins, favorites, and other data are also collected. After conducting a multidimensional linear analysis, we evaluate the Durbin-Watson test, VIF, and R^2 to determine the accuracy of fit. Our findings indicate that collecting videos and playback from the previous day positively impacts playback for the following day. However, pursuing likes andcoins excessively is susceptible to restriction by the system.