Analysis On The Influencing Factors Of Online Live Broadcast Reward Behavior

As a new type of social platform,live broadcasting has shown a trend of blowout development in recent years,and attracted a large number of content producers(anchors)and content consumers(users)to participate in it.Now live broadcasting has become an important channel for anchors and users to interact in real time on the Internet.Now the network live broadcasting industry is very popular,and many popular anchors have made good profits.However,when the industry is developing,people in the society have a great prejudice against the network live broadcasting.For example,some people even say that this new industry has no future,but it also affects many people,so they need to completely strangle the industry.And there are many people who think that webcast is just something that some unemployed youth and some gangsters are bored to do.In fact,it is not the case.Many people who have studies,jobs and status are also paying attention to the live broadcast on the Internet.Many famous university students also watch the live broadcast after class.Some staff members go home from work to watch the live broadcast.The live broadcast on the Internet gradually enters people's lives and becomes a leisure way after people's spare time,rather than being strangled by the boring industry.With the rapid development of webcast industry,the research on the influencing factors of reward behavior and the prediction of users' reward will also play a key role in the anchor and even the platform.This paper selects the real desensitization data of Yingke live broadcasting platform,aims to put forward the marketing direction and suggestions for the platform through the analysis of the influencing factors of users' reward behavior and the classification prediction of reward users.This paper analyzes and models the data based on R software.In the analysis of influencing factors,we mainly use the logistic regression method to test the significance of each variable.At the same time,we use the "ggplot2" package of R software to draw.The main factors that affect the live broadcast reward behavior are: the length of users' live viewing,the number of users' likes,the number of users' comments,and the average daily life of users in 30 days The main methods used in model prediction are logistic regression,random forest and support vector machine(SVM).In the logistic regression method,the ten fold cross validation method is used to verify and improve the model;in the random forest method,the minimum error cycle method is used to select the optimal parameters(mtry and n Tree)to fit and verify the model;in the support vector machine method,different kernel functions are used to fit the data,the optimal function is selected to determine the model,and finally the AUC value is obtained By comparison,logistic regression is the best.Finally,according to the conclusion of the paper,the factors are analyzed,and the marketing and push suggestions for the live platform are put forward.