Detection of Individual and Collaborative Fraud in Financial Systems Using XGBoost

Abstract

Financial fraud in cooperative and financial systems has become increasingly complex, particularly with the emergence of collaborative fraud involving multiple users. Traditional fraud detection methods, which primarily rely on transaction-based analysis, are often insufficient to identify such coordinated activities. This study proposes a machine learning-based approach using Extreme Gradient Boosting (XGBoost) to detect both individual and collaborative fraud by integrating transaction data, user behavior, and interaction-based features.The dataset used in this study consists of financial transactions and user activity logs, which are preprocessed and transformed into relevant features, including behavioral patterns and user interaction indicators. A multi-class classification model is developed to categorize activities into normal behavior, individual fraud, and collaborative fraud. The performance of the proposed model is evaluated using accuracy, precision, recall, and F1-score metrics, and compared with a baseline model that utilizes only transaction-based features. The results show that the enhanced model significantly outperforms the baseline model, achieving higher accuracy and improved detection capability, particularly in identifying collaborative fraud cases. The inclusion of behavioral and interaction features proves to be effective in capturing coordinated fraudulent patterns that are not detectable through transaction data alone.This study contributes to the advancement of fraud detection by introducing a comprehensive approach that considers both individual and collaborative behaviors. The proposed method provides practical value for financial institutions, especially cooperative systems, in improving fraud detection accuracy and strengthening internal control mechanisms. Future work may focus on real-time implementation and the integration of explainable artificial intelligence techniques to enhance model transparency.