The Clustering And Causal Inference Analysis Of Livestreamers Based On Live Streaming Data

Live e-commerce is a subdivision of online live broadcasting.With the entry of Tik Tok and Kuaishou and other social content platforms with advantages of huge users,the market size of live e-commerce has expanded rapidly.In 2020,an epidemic came out of nowhere.Long-term working and staying at home accelerated the formation of consumers’ habits of shopping through live-streaming,and live e-commerce ushered in the second high-speed growth.As of June 2021,the scale of users with live e-commerce reached 384 million,and grown continuously.On the other hand,due to the low threshold and considerable income,people from all walks of life have poured in,grassroots,celebrities,Internet celebrities,corporate executives,and even CCTV hosts and government officials have also entered the live broadcast room,forming a kind of phenomenon that everyone can be the livestreamer.In fact,the data shows that the live broadcast effect between e-commerce anchors is very different.Therefore,exploring the characteristics of livestreamers and making causal inferences about their live streaming effects will not only help the livestreamers to improve their own live streaming capabilities,but also provide feasible solutions for brands and MCN institutions to select livestreamers to achieve their own marketing promotion and profit goals.It is of great significance to stabilize economic development under the normal state of the epidemic.Based on live streaming data,this paper takes e-commerce livestreamers as the main research object,and clusters the livestreamers from six aspects: personal characteristics of livestreamers,characteristics of livestreamers activity,characteristics of live streaming popularity,characteristics of live streaming interaction,live streaming commodity characteristics,and live streaming audience characteristics with factor analysis,and analyzes the characteristics of different types of livestreamer groups and the differences in live streaming effects.Then,according to previous research and cluster analysis results,“whether the livestreamer signs a contract with MCN institution” and “interactivity” were selected as processing variables respectively,and the causal relationship between the above processing variables and the live streaming effects of the livestreamer was verified through propensity score matching.The results of the empirical analysis show that: First of all,whether the livestreamer signs with MCN institution,the type of the livestreamer,and the 10 factors after dimensionality reduction all have a relationship with the average sales and the average conversion rate.Among them,the livestreamer type,whether the livestreamer signs with MCN institution,the livestreamer popularity,livestreamer,activity and live interaction have the highest contribution to the clustering results.Secondly,through cluster analysis,the livestreamers are divided into three categories: popular,active and potential.There are significant differences in the characteristics and live streaming effects of the three types of livestreamer.Finally,the causal inference analysis of the live streaming effects of the livestreamers shows that the livestreamer signing with the MCN institution has a significant role in promoting its sales and the conversion rate.On the other hand,since sales are affected by the number of viewers,interactivity has a reverse impact on the average sales and the conversion rate.The conversion rate first increases to a certain extent and then gradually decreases.