Telecom customer churn prediction method based on cluster stratified sampling logistic regression

This paper provides a novel and efficient method for predicting potential customer churn from imbalanced data set of Orange Telecom and UCI. Customer churn is always a rare event, but it is necessary to be paid attention. The main intended contribution of this paper is to apply binary logistic regression model (LRM), which seldom be used in the problem of imbalanced data prediction. The parameters estimated for the data gathered with serious problem of imbalance, therefore we take stratified sampling method, and improve traditional logistic regression model parameters estimated methods. The experimental results show that our prediction method performs satisfactorily, and it can be effective to forecast the telecom customer churn.