Helpfulness Modeling And Prediction Of Short Video In Social E-Commerce

In recent years,China’s social e-commerce platforms have developed rapidly,among which short video marketing has played an important role.Because of its convenient communication and strong interaction,this marketing method is favored by more and more users.As the platform of short video management,how to judge whether the video is "helpful" for users is an urgent problem to be solved,because "helpful" videos are an important basis for the platform to make accurate recommendation.At present,a large number of studies focus on the helpfulness of text and pictures in online comments,but few on the helpfulness of short videos.In view of this,this dissertation takes short videos in social e-commerce platforms as the research object and conducts an indepth study on the helpfulness of short videos based on information adoption model.In theory,the influencing factors of the helpfulness of short videos are sorted out from the two aspects of short video content and short video publisher,and a theoretical model of influencing factors of short videos is established.The significant influencing factors are found by using panel random effect regression method.Based on this,a short video helpfulness prediction model was constructed by using machine learning stacking integration method.In terms of methods,short video data are difficult to measure and individual effect is difficult to understand.Text emotion and audio emotion are difficult to measure.The text emotion of video description is calculated based on the long short-term memory networks,and the audio emotion of video content is calculated based on the mel frequency cepstrum coefficient.For the problem that individual effect is difficult to understand,the intercept term of individual heterogeneity is introduced into panel random effect method in the model solving stage.The empirical research of little red book shows video description text readability,video publishers action,the emotional value of the video,video comments,video publishers fans number,the number of likes accumulated by the video publisher and total number of videos released by video publishers all have large effects on the helpfulness of a short video.The empirical research results based on the constructed prediction model show that the proposed method is effective in predicting the helpfulness of short videos.In this thesis,the research on the helpfulness of short videos in social e-commerce platforms enriches the previous information types,extending from text comments and pictures to videos,and perfecting the theoretical model of helpfulness of online comment.The proposed short video helpfulness prediction method provides a theoretical basis for the accurate recommendation of short video.