查看文章

In today's competitive conditions, the importance of minimizing costs is increasing day by day. As a result of the researches, it has been determined that the cost of attracting new customers is 10 times more than the cost of holding existing customers. This increases the importance of customer churn analysis, too. In this review, machine learning algorithms such as artificial neural networks (ANN), decision trees, support vector machines (SVM), naive bayes, k-nn and extreme gradient boosting (XGBoost) and customer churn analysis, Cox proportional hazard model and deep learning techniques. In addition, customer churn analysis studies conducted in various sectors using these techniques were examined. When the customer churn analysis studies are examined, it is seen that more complex systems can get modeling and reached higher success rates with deep learning technique. However, because it is a new technique and could not give stable results, machine learning algorithms, which are the closest alternative, are thought to be able to useful in estimating time-related events such as customer churn. The Cox regression model was found to be successful in estimating the independent variables affecting the time variable, the rate of life expectancy and the groups under risk. It is seen that deep learning techniques give better results in more complex structures.