Multi-Perspective Financial Fraud Detection Using Integrated Learning Models and Cloud Computing Platforms

View at publisher Summarise with VIRAT
Authors: Sathyamoorthy P, Lenin Raj S.J, N. Kathirvel
Cited by: 0
References: 15

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Financial transaction fraud is one of the most serious challenges for modern digital payment systems, causing enormous monetary losses all around the world. This work presents a comprehensive fraud detection study involving several machine learning techniques operating on large-scale datasets comprising five million financial transactions and then considering various models, such as XGBoost, MLP, Random Forest, and Isolation Forest, on a highly imbalanced dataset, in which fraudulent transactions make up only about 3.59% of the total. Among the models considered, XGBoost returned the best result, with an AUC-ROC of 0.9912 and an Average Precision of 0.9734, indicating a high capability in capturing complex and subtle patterns of fraud. The importance of feature engineering is also underlined, since it has been demonstrated that transaction velocity, temporal behaviors, and network-based features increase the accuracy of detection considerably. In this respect, these findings are very useful for constructing more reliable and efficient fraud detection systems for real-world financial environments.

Fields of Study & Subjects

Imbalanced Data Classification Techniques Privacy-Preserving Technologies in Data Benford’s Law and Fraud Detection

Publication details

Journal / source
International Conference on Intelligent Sustainable Systems (ICISS)
Pages
1332–1337
Publication date
4 March 2026
Language
EN