Purpose

The high increase in the number of companies competing in mature markets makes customer retention an important factor for any company to survive. Thus, many methodologies (e.g., data mining and statistics) have been proposed to analyze and study customer retention. The validity of such methods is not yet proved though. This paper tries to fill this gap by empirically comparing several techniques: Customer churn - logistic regression models, Random forest, Gradient boosting machine and Deep neural network. The paper proves the superiority of logistic regression technique and stresses the needs for more advanced methods to churn modelling.
Let’s consider a typical engagement journey in a Telecom Company.