Evaluation of CART and XGBoost Methods on Customer Loan Risk Prediction Based on Consumer Behavior

The demand for money lending services has surged significantly in Indonesia, with financial institutions such as banks, cooperatives, and other lenders striving to offer faster and more accessible loan application processes to attract customers. However, determining whether a customer is eligible for a loan requires thorough due diligence to ensure smooth repayments and minimize potential losses for the lender. The process of approving or rejecting a credit application is often complex and time-consuming. This study aims to address the challenge of predicting customer credit eligibility by employing two machine learning techniques: Classification and Regression Tree (CART) and eXtreme Gradient Boosting (XGBoost). The research follows a structured methodology, including data acquisition, pre-processing, splitting the data into training and testing sets, applying the CART and XGBoost algorithms, and evaluating the models' performance. Through this approach, the study seeks to enhance the accuracy and efficiency of credit approval decisions, helping financial institutions streamline their processes.