In the telecommunication industry customer classification and loss analysis

Nowadays, many companies organize many studies and campaigns to prevent the loss of their existing customers and to find ways to regain their customers who are about to be lost, rather than trying to acquire new customers. Various surveys are very popular nowadays, especially in the telecommunications industry where it is easy to lose customers. In the telecommunications sector, in order to avoid losing customers, the profiles of lost customers are determined and studies are carried out as a result of the information obtained from the study on those who have not yet been lost, thus preventing the loss of existing customers. In this study, the data set published on kaggle.com was used. Five classification methods, which are data mining classification models, were used to find the reasons for losing customers with a set of demographic data belonging to a telecommunications company. In the study; Naive Bayes, Decision Tree, Logistic Regression, Random forest and KNN were used. These classification techniques have been tried in 4 different scenarios. These scenarios are: creating a model by including all independent variables, creating a model with the 3 least effective independent variables removed, creating a model with the 5 most effective independent variables, and creating a model by removing customers who have been registered for 6 months or less. Among these modeling methods used, the method that can make the best prediction is Logistic Regression modeling, which is evaluated with the precision, accuracy, ROC curve, F1 score and Recall used in the evaluation of the results, and is done with customers who have been registered for 6 months or less among these scenarios, due to its superior performance compared to other models. has had good performance.