Data Mining and Analysis of Video Barrage By AI Algorithm

Abstract

The barrage is an important form for the audience to express their emotions and opinions. It runs through the entire video and feeds back audience's overall evaluation of the plot type, characters, and even actors of the videos. Mining such information from massive barrages not only has important academic value but also provides a reference for relevant business decisions to increase film traffic and revenue. We crawled the bullet screen information of 5 different types of recommended movies in the Bilibili bullet screen network in 2021 and performed statistical analysis on the data in a graphical visualization manner. The user's attention is analyzed through the word cloud map. The distribution of the number of bullet screens was used to reflect the changes in the number of viewers every week and every day, and the degree of attention during the film screening process was analyzed. Sentiment analysis is performed on the obtained bullet screen data based on artificial intelligence algorithms. First, the Word2vec model generated the word vector of the bullet screen text and input it into the machine learning model SVM and the deep learning model TextCNN for classification. The experimental results show that the deep learning model is higher than the traditional model in accuracy.