Bank Customer Minimization Churn Rate Using Machine Learning

Customer churn poses a significant challenge for banks, affecting revenue and customer satisfaction. This study employs machine learning techniques, specifically Naive Bayes, Decision Tree, and AdaBoost classifiers, to predict and minimize churn rates. The process involves thorough data preprocessing, feature engineering, and encoding, addressing issues such as duplicate values, missing data, and categorical features. Exploratory data analysis (EDA) and visualizations provide insights into the distribution of customer attributes. Undersampling using the NearMiss algorithm is applied to balance the dataset, and the models are evaluated using essential metrics like recall and ROC-AUC score. Hyperparameter tuning is conducted to optimize model performance. The chosen AdaBoost model demonstrates superior recall on both the training and test datasets, making it the preferred model. The evaluation extends to the ROC-AUC curve, illustrating the model's trade-off between true positive rate and false positive rate. The final model's predictions are exported, and the results showcase the actual and predictedchurn status of customers. This comprehensive approach aims to equip banks with an effective tool to proactively identify and retain customers, ultimately mitigating churn and enhancing overall performance. IndexTerms: Bank Churn Prediction, Customer Retention, Churn Analysis, Credit Card Churn, Banking Sector, Customer Attrition, Financial Services, Supervised Learning, Machine Learning.