Forecasting Fraud Detection Using Data Science Methods

Abstract

Fraud detection is critical in various domains, including finance, healthcare, and e-commerce, where fraudulent activities pose significant threats to organizational integrity and financial stability. Traditional fraud detection methods often fail to address the dynamic nature of fraudulent behavior. In response, data science methods have emerged as promising tools for forecasting fraudulent activities by leveraging advanced analytics techniques on large-scale datasets. This research will make significant contributions by focusing on predicting fraud detection through data science methods. The findings will guide on preventing customers from committing fraud. The research questions aimed to be answered in this study are as follows: What are the key factors affecting fraud detection? Which customer behaviors are the strongest predictors of fraud detection? This study will provide a valuable model to the industry, enabling financial institutions to strengthen their risk management strategies and translate innovations in AI into applications.

Keywords

• Fraud detection
• Logistic regression
• XGBoost classifier
• CatBoost classifier
• Random forest

References

1. Kavus, B., & Soleimani-Zakeri, N.S. (2024). Forecasting fraud detection using data science methods. The Eurasia Proceedings of Science, Technology, Engineering & Mathematics (EPSTEM), 31, 1-10.