Research And Implementation Of User Analysis Model Based On Short Video Platform

The Internet and multimedia news industry are developing rapidly.The short video platform on the mobile terminal has gradually entered the era of popularization,and the short video platform has gradually become a trend of development.Due to the short time,rich content,and the ability to watch anytime,anywhere,it has gradually occupied the fragmented time of users.The exponential growth of users has soared,and the huge user traffic has attracted the interest of the e-commerce industry and the advertising industry.Through video anchors to promote products in the form of cooperative endorsements,a new profit model has formed.However,the number of users exceeding 100 million will cause difficulties for the e-commerce and advertising industries in finding cooperative users.Therefore,if users with a larger user influence can be screened out from a large number of users,or if it is possible Potential users of e-commerce cooperation can save a certain amount of time or cost.This paper aims at this point of view,and has certain practical significance for studying the rapid integration of the short video industry and the e-commerce industry.This paper first describes the current research status of user influence calculation in social networks,introduces the In-Degree algorithm based on the number of fans,and the association algorithm based on the Page Rank algorithm,and points out some of their shortcomings,such as the Page Rank algorithm,fans evenly distribute the PR value to the concerned users,which is unreasonable.In view of the shortcomings of the above algorithms,the SF-UIR influence algorithm based on the Weibo network platform was introduced,and the application and implementation of the SF-UIR algorithm in the short video platform was discussed.We compared the ranking of the influence received by the user with the ranking of the user in the factor analysis scores through experiments,and discussed the impact which the user's platform certification label and the quality of recent works brought,and we found the SF-UIR algorithm can be applied to short video platforms.Subsequently,in order to generate a user analysismodel that can judge the potential of the user's e-commerce,a multi-label learning method is introduced.to explain the significance of multi-label learning.Compared with single-label learning,multi-label learning can grasp the rules and characteristics of things more comprehensively.Subsequently,two methods of multi-label learning,a problem conversion method and an algorithm conversion method are explained.Point out some of the limitations in the problem conversion method.As for the algorithm conversion method that improves the single-label algorithm and incorporates into multi-label learning,the implementation principles and usage strategies of ML-KNN,kernel-based SVM,XGBoost model were described.Subsequently,we introduced the Network construction process based on multi-label learning of multi-layer BP neural network.Through experiments comparing the performance of the above algorithm model on the user data set of Tik Tok,it was found that XGBoost performed better,and can be selected as the user analysis model we need.In the experimental phase of the user analysis model,the SF-UIR influence algorithm is combined with the user analysis model to identify potential e-commerce cooperation users.The user analysis model is used to discover potential high-quality video authors who have the possibility of becoming authenticated users.Combining the two parts of the user mining experiment analysis,it is believed that the user analysis model has a certain user influence discrimination and potential e-commerce user discovery ability,and also believes that the user analysis model has a certain discovery ability for potential high-quality video authors on the short video platform.However,because the diversified factors of the official certification of the short video platform cannot be overcome,there are still some shortcomings in the analysis of the authors in some professional fields.At the end of this article,the implementation of the data crawler for the short video platform based on Fiddler4 and the construction process of the user data set are introduced.