Optimization of Machine Learning Algorithms for Fraud Detection in E-Payment Systems
Open Access
- 17 February 2025
- journal article
- Published by iLearning Journal Center in Journal of Computer Science and Technology Application
- Vol. 2 (1) , 55-64
Abstract
This study explores the optimization of machine learning algorithms for fraud detection in electronic payment (e-payment) systems. The rapid growth of e-payment platforms has introduced significant challenges in ensuring the security and integrity of financial transactions. Fraud detection plays a pivotal role in mitigating these risks, and the application of machine learning (ML) has emerged as a powerful tool to identify fraudulent activities. This research examines how Data Quality (DQ), Algorithm Selection (AS), and Optimization Techniques (OT) influence Model Performance (MP) and, subsequently, Fraud Detection Effectiveness (FDE). The study utilizes Partial Least Squares Structural Equation Modeling (PLS-SEM) through SmartPLS 3 to analyze the relationships between these variables. The results demonstrate that high Data Quality significantly enhances Model Performance, while Algorithm Selection and Optimization Techniques also contribute positively, albeit to a lesser extent. The findings reveal that Model Performance plays a crucial mediating role between these factors and the effectiveness of fraud detection. Fraud Detection Effectiveness is found to be significantly impacted by Model Performance, suggesting that improving model accuracy and efficiency is essential for better fraud detection outcomes. Reliability and validity tests show strong internal consistency for all constructs, with Cronbach’s Alpha, Composite Reliability, and Average Variance Extracted (AVE) all reaching satisfactory levels. The study highlights the importance of data preprocessing, the careful selection of machine learning models, and optimization techniques in achieving high-performing fraud detection systems. The results provide valuable insights for the development of more robust and scalable fraud detection mechanisms in e-payment systems, contributing to the broader field of machine learning and cybersecurity. Future research could explore advanced techniques like deep learning and blockchain integration for further enhancement of fraud detection systems.Keywords
This publication has 10 references indexed in Scilit:
- Optimizing Blockchain Digital Signature Security in Driving Innovation and Sustainable InfrastructureBlockchain Frontier Technology, 2025
- Ethical Framework for Artificial Intelligence and Urban SustainabilityBlockchain Frontier Technology, 2025
- The Role of Machine Learning in Improving Robotic Perception and Decision MakingInternational Transactions on Artificial Intelligence (ITALIC), 2024
- AI for Sustainable Development: Applications in Natural Resource Management, Agriculture, and Waste ManagementInternational Transactions on Artificial Intelligence (ITALIC), 2024
- E-Commerce Fraud Detection Based on Machine Learning Techniques: Systematic Literature ReviewBig Data Mining and Analytics, 2024
- Detecting Frauds and Payment Defaults on Credit Card Data Inherited With Imbalanced Class Distribution and Overlapping Class Problems: A Systematic ReviewIEEE Access, 2024
- Defending against digital thievery: a machine learning approach to predict e-payment fraudInternational Journal of Management Practice, 2024
- Strategy Based Technology-Based Startups to Drive Digital Business GrowthStartupreneur Business Digital (SABDA Journal), 2023
- Peer to Peer Lending Business Agility Strategy for Fintech Startups in the Digital Finance Era in IndonesiaStartupreneur Business Digital (SABDA Journal), 2023
- NUS: Noisy-Sample-Removed Undersampling Scheme for Imbalanced Classification and Application to Credit Card Fraud DetectionIEEE Transactions on Computational Social Systems, 2023