<![CDATA[Customer churn prediction is a crucial aspect of the banking industry for maintaining customer loyalty and reducing the cost of acquiring new customers. This study aims to develop a web-based decision support system capable of predicting potential customer churn using the Gradient Boosting Machine (GBM) algorithm. The dataset used is the Bank Customer Churn Dataset, consisting of 10,000 customer records with 14 attributes. The research stages include exploratory data analysis and preprocessing, which involves data cleaning, categorical feature encoding, feature engineering (BalanceSalaryRatio, TenureByAge, CreditScoreGivenAge), and data balancing using SMOTE to address class imbalance. The GBM model was trained on the balanced dataset and evaluated using accuracy, precision, recall, and F1-score metrics. The evaluation results show that the model achieved an accuracy of 83.95%, with a recall of 67.32% for the churn class, indicating a strong capability in identifying customers at risk of churn. Feature importance analysis reveals that Age and NumOfProducts are the most dominant features, contributing approximately 77% to the prediction. The model was then implemented in a Flask-based web application with an HTML and CSS interface, enabling non-technical users to perform real-time churn predictions. This system is expected to assist banking institutions in designing more targeted and data-driven customer retention strategies.]]>
Copyrights © 2026
