Analyzing And Predicting The Influences Of E-commerce Celebrities' Sales Based On Data Mining

With the rise of social software,live streaming platform and we media platform,the forms of Internet content consumption are becoming more and more abundant.From early forms of text and pictures to real-time video streaming,the forms of online marketing are gradually diversified.Since 2017,with the rise of live streaming and short video,platforms,MCN institutions and free creators and other parties have been able to develop rapidly and gradually extend to the vertical field.E-commerce celebrity refers to a "network celebrity" who carries out marketing activities through social platforms and finally achieves a deal on e-commerce platforms.According to the prediction of clauri,a domestic research institute,the e-commerce market in web celebrity economy will reach 300 billion yuan in 2020.To explore and study the behavioral patterns and characteristics of e-commerce celebrities on social platforms,and to excavate the potential relationship between the business value of ecommerce celebrities and their behaviors on social platforms is conducive to a deeper understanding of this new economic phenomenon,so as to evaluate the business value of e-commerce celebrities and improve their social marketing methods.At present,most of the domestic researches on e-commerce Internet celebrities are based on individual case studies and qualitative analysis.Therefore,this paper takes Douyin platform as the specific measurement object,establishes the original data set of e-commerce online celebrity.Based on this data set,it analyzes the business behavior and attribute information of e-commerce online celebrity,establishes the business value model of e-commerce online celebrity,and evaluates and predicts the business value of e-commerce online celebrity.First of all,this paper comprehensively USES the crawler technology and thirdparty monitoring platform information to obtain the original data set,which includes the full amount of behavior data of 102 e-commerce celebrities on Douyin platform,as well as the marketing behavior characteristics and sales data between March and July2020.Secondly,based on the data set of e-commerce online celebrities,the typical marketing behaviors of e-commerce online celebrities are extracted and analyzed,including the behaviors of carrying goods,live broadcasting and topic.Mining the statistical laws of different behaviors,based on the behavioral analysis to build ecommerce red marketing behavior characteristics,daily behavior characteristics,basic information characteristics,a total of 41 characteristic attributes.Thirdly,on the basis of the above feature engineering,the random forest?K-NN classification algorithm and Logistic regression are used to construct the e-commerce online celebrity sales level evaluation model based on different features.The experimental results show that the highest accuracy rate of the model is 0.8.Meanwhile,10 characteristics that can most affect the sales volume of e-commerce celebrities are extracted.