Research Of Telecom Customer Churn Prediction At Different Time Periods

With the gradual saturation of the telecom industry in recent years,the competition among the three major operators in the country has become increasingly fierce,resulting in high customer churn rate,reducing customer churn becomes a major problem that operators need to deal with.According to research,the cost operators developing a new customer is six times that maintaining an old user,and high-value old customers bring significantly higher profits to the company than new customers,so maintaining customers and reducing churn is the focus of operators..New users who have just entered the network are more likely to lose than older users.There are also differences in the factors affecting the loss of the two types of users.Therefore,this paper focuses on the customer retention time,and studies the user loss problem in both short-term and long-term aspects.For the short-term loss of users,the focus is on accurately predicting users with off-network tendencies,shortening the forecast period,and adopting timely retention measures.For long-term users,pay attention to their retention time,risk probability,then implement continuous policy to maintain customers.The main work of this paper includes:(1)analyzing the current situation and research methods of customer churn,and exploring different perspectives of problem research.(2)In terms of data preparation,obtain short-term data and long-term data respectively from a telecom operator,design combined features according to the daily and monthly indicators,then preprocess data.In terms of feature selection,use different feature selection methods according to the characteristics of models.(3)For short-term customer churn,select four data mining methods to construct shortterm loss warning model,such as logistic regression,support vector machine,random forest,lightGBM,then evaluate and compare methods.(4)For long-term customer churn problem,the survival analysis method is used to construct the model,which focuses on analyzing the influencing factors?network duration probability of loss,and the probability of the customer's off-network risk,and provides data support and theoretical advice for the operator to take retention measures.