Researches On The Analysis Of Live Streaming Behavior Based On Data Mining

In recent years,with the development of information network,the popularity of live streaming is getting higher and higher.In the process of live streaming,there are rich interactions between the broadcasters and the viewers,including the interaction of posting comments and buying gifts.The interaction between the broadcasters and the viewers promotes the viewers to integrate into the production of video content,not just the consumers of video content.By studying the interaction between the broadcasters and the viewers,we can not only understand the content production in the process of live streaming,but also promote the operation and development of the live streaming industry.This paper takes Longzhu live platform as an example to analyze the user behavior of the live platform in detail.Using the collected one-month dataset,the paper performs data modeling analysis on the live platform of Longzhu,from the perspective of the broadcasters and the viewers.It mainly includes the following aspects:Firstly,this paper studies the operation of Longzhu live streaming platform,focusing on the analysis of the live streaming type and the income distribution based on virtual gifts,and finds that the income distribution of the live broadcast type follows the stretched exponential distribution.The number of online broadcasters and the viewers has a high correlation with the income.Secondly,this paper focuses on the analysis of user behavior.Through the analysis of the number of live broadcast,post comments and gift distribution,it is found that post comments and gift distribution obey the heavy tail distribution,and the viewing number distribution not only obeys the heavy tail distribution,but also obeys the long tail distribution.Statistics show that the majority of male broadcasters are mainly distributed in the first-line cities.By clustering the distribution of the number of audience in a day,we find that the viewers can be divided into three groups.By analyzing the number of post comments,the number of gifts and the distribution of income ability,it is found that the number of post comments,the number of gifts and the distribution of income amount all present heavy tailed distributions.Thirdly,through the construction of the relationship network of the broadcasters,the similarity of the nodes is introduced into the module degree function to discover the community of the broadcasters group,and four types of broadcasters groups are obtained through further statistics.Then the integrated learning model is used to predict the income of the high-income anchor group,and the accuracy of the model is verified by experiments.