Research On The Prediction And Analysis Model Of Mobile Network User Loss Of J City Chinaunicom Company

With the development of communication technology,the number of mobile network business users is growing rapidly,and the market competition of communication business is becoming increasingly fierce.In the case that new users are difficult to improve,in order to ensure the sustainable development benefits of communication service enterprises,how to maintain the stock of users,how to identify the loss tendency of users in advance,how to formulate measures to effectively reduce the loss rate of users and so on have become the key problems that operators need to solve in the next development.Taking the mobile network business users of J City Unicom Company as the research object,this paper makes statistical analysis on the loss of mobile network business stock users of the company in 2018 and excavates the reasons for the loss.It is learned that the customer service department needs to screen the pre loss users of the whole network with a large workload and a low accuracy,the maintenance effect is not ideal,and the loss rate of users is still high,We need to design a scheme to improve the maintenance effect of stock users.Then this paper uses the methodology of cross industry data mining standard process CRISP-DM and logistic binary regression algorithm to design the user churn prediction scheme.Through the correlation test method between the independent variable and the dependent variable,the network duration,preferential activity participation identification,total consumption,data traffic usage,calling call duration,called call duration,and the number of complaints are screened out There is a significant correlation between the characteristic indexes and the state of user churn.The model expression lost(logit(P))= 0.203 × on-line time + 0.012 × preferential activity participation ID-1.191 × total consumption-0.068 × data traffic usage + 0.195 × calling call time + 0.032 × called call time + 0.553 × complaints-1.338 is constructed.The overall prediction accuracy of the model can reach about 86% ? Finally,through the application of the established model,the paper puts forward the optimization scheme of dividing the maintenance target user group according to the prediction results of the model and the suggestion of refining the maintenance measures according to the model characteristic index,summarizes the analysis scheme of the loss prediction model to carry out the stock user management,which can improve the maintenance efficiency of the mobile user management and reduce the loss rate of the mobile user in J City Unicom It has good effect.