Kredi Kartı Müşteri Kaybının Makine Öğrenmesi Yöntemleri Kullanılarak Tahmin Edilmesi

Today, customers are giving up using credit cards for various reasons and this has negative consequences for banks. Therefore, it is necessary to predict potential customers who will cancel their credit cards in advance and to turn these cancellations in favor of the bank and thus to regain the customers. This situation is also very important in terms of monitoring customer loss and preventing such loss. In this context, a model is proposed using machine learning methods to detect the card cancellation status of customers using credit cards and thus predict customer loss. A dataset obtained from the Kaggle platform was utilized to create the model. This dataset contains credit card data belonging to a total of 10127 customers. Although there were 23 features in the dataset, 2 features were deleted without being included in the model because they did not affect the results. As a result, a total of 21 different variables were used, 20 inputs and 1 output. The models were created using Artificial Neural Networks, Logistic Regression, Support Vector Machines, K-Nearest Neighbor, Decision Tree, Random Forest, Ada Boost, and Gradient Boosting machine learning algorithms. As a result, it was seen that the model with the highest performance was Gradient Boosting with a rate of 98.70%, and the model with the lowest performance was Support Vector Machines with a rate of 67.9%. All these results clearly show that Credit Card Customer Churn can be effectively predicted by machine learning methods.