Comparison of logistic regression and decision tree for customer churn prediction in Telecommunications

Customer churn, loss of customers due to switch to another service provider or non-renewal of commitment, is very common in highly competitive and saturated markets such as telecommunications. In order to solve this problem, predictive models need to be implemented to identify customers who are at risk of churning and also key drivers of churn need to be identified. In this study, two models for prediction of customer churn in next 45 days are compared - logistic regression and decision tree. The dataset used contain 16 variables and 50,000 customers in both training and testing data set. Decision tree outperformed in predictive performance logistic regression with hit rate 81.1% and specificity 94%. The most important variables in both classification models were customer duration and contract duration and in logistic regression model also value added services played a big role.