Research Of Cellphone Users' Replacement Behavior Based On The Big Data Mining

With the rapid development of Internet technology and the popularization of computer application,data acquisition ability of the cellphone industry gets unprecedented increase,obtaining and accumulating a voluminous moving-communication data.Through big data mining technology,making the research on the replacement characteristics of cellphone users and affecting factors of replacement behavior is of great significance,it can enable us to deeply grasp the user's needs,and then achieving exact-marketing.Based on this background,this paper puts forward the research of cellphone user's replacement behavior based on the big data mining.First,relying on the subsystem of the provincial telecom operators,obtains a certain period data of cellphone users,on the basis of defining the connotation of the cellphone users replacement behavior and building the index system,draws out the samples which Kullback-Leibe information reaches more than 0.99995 according to simple random sampling method,and does much work including the management of data,the handing of missing values,the detection of the outliers,ensuring the quality of samples;afterwards,this paper carries the exploratory analysis to the data,shows the user characteristics,phones brand characteristics and phones performance characteristics before and after replacement through visualization technology;then,respectively uses the Decision tree,Adaboost,Bagging and Random forest algorithms for training data,establishing prediction models,getting the influence of variables on replacement behavior,compares the different models according to the average misjudgment rate of cross-validation;finally,summarizes various models,combining with the fact,concludes the influence of the variables on the replacement behavior,and puts forward the recommendation that the cellphone manufacturers should pay more attention to the brand construction,increased investment in technology and innovation dynamics.This paper bases on the real data,uses research methods combining the qualitative and quantitative analysis;makes an in-depth research on the replacement behavior through visualization technology,on this basis,analyzes the factors affecting replacement behavior through introducing many algorithms and puts forward some suggestions on production and sales,it has certain innovation.However,there are still some remains to be improved on index selection,they can be improved in the future research work.