Customer retention in commercial banking as a classification task in machine learning

Abstract

Introduction. Customer churn is a common problem for many industries, particularly the banking sector. To thrive, banks need to attract new customers, as each lost customer leads to a decrease in profit and requires time and effort to acquire a new one. Customer churn occurs when a client ceases to use a bank's product or service. Retaining customer interest is more beneficial and cost-effective than attempting to attract new ones. Therefore, reducing customer churn becomes one of the key tasks for businesses. Banks that can retain and attract new customers have significantly higher chances of success. Hence, the use of machine learning methods becomes one of the key tools for addressing the task of reducing customer churn. These methods have the potential to help banking institutions optimize their processes and increase profitability. Purpose. The aim of the study is to assess the effectiveness of using machine learning methods for customer retention in a bank, including their construction, testing, and evaluation of the economic impact. Method (methodology). This article investigates the issue of retaining customers of a commercial bank by determining the probability of customer churn using classification methods of machine learning. Logistic regression models (GLM), decision trees (Decision Trees), random forests (Random Forest), as well as support vector machines (SVM), k-nearest neighbors (k-NN), and naive Bayes algorithm will be constructed for this purpose. The quality of the constructed models will be evaluated using a confusion matrix. Results. The obtained results revealed high accuracy of the constructed models and their ability to effectively identify bank customers prone to churn. The conclusions of this article may be valuable for developing customer retention strategies not only for commercial banks but also for various business sectors where customer attrition is a relevant issue.