Customer separation estimation with gradient boosting algorithm for machine learning

Customer churn can significantly affect a business's profitability; therefore, it is important to develop strategies to predict and prevent these separations in advance. Techniques and algorithms aimed at this goal provide deep insights into customer behavior and trends, helping companies manage customer churn more effectively. The aim of this study is to improve the ability to predict customer churn by using various data mining techniques and machine learning algorithms. Another objective is to discuss how these machine learning algorithms can be applied in the telecommunications and banking sectors and how they can be used to reduce churn. In this study, two datasets have been used. The telecommunications dataset contains 7,043 data points with 21 different features, and the banking dataset contains 10,000 data points with 14 different features. 80% of the data in the datasets were used for training the model, and 20% were used for testing. The best hyperparameter tuning was applied to improve the performance of the Gradient Boosting algorithm, and after each parameter change, we observed its impact on the model's performance. GridSearchCV and RandomSearchCV was used for the best hyperparameter values. The SMOTE and SMOTEEN function were used for imbalanced data, and by adjusting the SMOTE parameters, the model's performance was enhanced.