XGBOOST-BASED FRAUD TRANSACTION CLASSIFICATION ANALYSIS IN ONLINE PAYMENT SYSTEMS

Abstract

The rapid development of online payment systems has significantly facilitated digital transactions; however, it has simultaneously increased the risk of fraudulent activities. Fraud detection has become a critical challenge due to the complex characteristics of transaction data and the imbalanced class distribution between legitimate and fraudulent transactions. This study aims to analyze the performance of the XGBoost algorithm in classifying fraudulent transactions within online payment systems. The research employs the Online Payments Fraud Detection Dataset obtained from the Kaggle platform. The research methodology consists of several stages, including dataset collection, data preprocessing, categorical data transformation using label encoding, feature engineering for the generation of new attributes, data partitioning through split validation with an 80:20 ratio, model development using the XGBoost algorithm, and performance evaluation using a confusion matrix, accuracy, precision, recall, F1-score, and Area Under the Curve (AUC). The experimental results demonstrate that the XGBoost model achieves excellent classification performance, with an accuracy of 99.98%, precision of 85%, recall of 100%, F1-score of 92%, and an AUC value of 0.9996. Furthermore, feature importance analysis reveals that errorOrig and newbalanceOrig are the most influential attributes in detecting fraudulent transactions. Based on these findings, it can be concluded that the XGBoost algorithm is highly effective for fraud transaction classification in online payment systems and exhibits strong potential for implementation in automated fraud detection systems to enhance the security of digital financial transactions.