Bank customer loss analysis with machine learning methods

Customer churn is the percentage of customers who stop using the company's product or service and terminate their relationship. It can be more difficult to gain a new customer than to lose one. Therefore, the company's loss is minimized when the existing customer is predicted to leave the business. Computer sciences such as artificial intelligence are used to predict customer defection. Machine learning methods are one of these fields. Machine learning is a branch of computer science and artificial intelligence that focuses on the use of data and algorithms by imitating the way humans learn in general. In this study, a ready-made publicly available data set was used to predict the customer churn of a bank taken from the Kaggle site. The dataset consists of 10,000 data and 14 categories. The data set includes bank customers from France, Germany and Spain. The dataset consists of variables including customer number, surname, country, gender, credit score, age, length of time as a customer, balance, number of products, whether they have a credit card or not, whether they have left or not, and numpy, pandas, seaborn, matplotlib libraries were used. In the study, machine learning classification algorithms such as Logistic regression, K-nearest neighbor, Decision tree, Random forest, LightGBM (light gradient boosting), Catboost (categorical boosting) classification algorithms were used to determine how to detect customer churn and the performances of the models were compared and the results were examined in detail.