Machine Learning Approaches for Bank Customer Churn Analysis

With the development of technology, rapid developments are experienced in the banking sector, as in many other sectors. However, with these developments, many problems arise, especially customer losses. Customer losses are a common problem not only in banks but also in many institutions such as insurance, production and logistics, etc. Customer churn is when people performing banking activities begin to abandon these services. It also causes financial and prestige loss for banks. It is possible to minimize or eliminate this problem by providing good service to customers. However, in order to achieve this, it is necessary to accurately analyze the trends that customers care about when providing banking services and the characteristics that cause customer loss. For these reasons, this study aims to estimate the probability of a bank customer leaving the bank from which he receives service. In this study, a data set containing customer data of a bank was used. The data set has various features that include customers' demographic information, financial situations, and interactions with the bank. Using common machine learning algorithms such as Logistic Regression, Decision Trees, Support Vector Machines (SVM), Gradient Boosting, K-Nearest Neighbors (KNN) and Multi-Layer Perceptron (MLP), the reasons for bank customers leaving were determined and the results were evaluated. Cross-validation method was used to evaluate the performance of each proposed machine learning algorithm. Additionally, hyperparameter optimization was used to increase the classification accuracy of the proposed methods. The best parameters were determined with the GridSearchCV hyperparameter optimization method. The performances of the proposed machine learning algorithms are compared. When the findings were evaluated, it was observed that the Gradient Boosting method achieved the highest accuracy rate.