Analysis on Bilibili Video Categorical Segmentation Using XGBoost

Video media platforms such as YouTube and Bilibili, have been increasingly popular. The industry of video media is developing rapidly and it is one of the main industries that bring monetary to many countries. Bilibili is one of the largest and most popular video sites in China. To make better policies and decisions as well as remain competitive, organizers and uploaders of Bilibili have to understand the features of popular videos. Marketing segmentation can help to achieve this goal. This study presents Bilibili video segmentation with XGBoost, an advanced machine learning algorithm based on tree boosting system. The study was divided into two sections: the categorical analysis and the experiment. The statistical analysis and the experiment are all based on data of videos of the 100 most popular uploaders. In the experiment, the multiclass log loss of XGBoost was 0.519707. The results of this study will be helpful for the management and predictive analysis of Bilibili and the whole video media industry.