Artificial Intelligence and Machine Learning–Based Fraud Detection Frameworks for Real-Time Transaction Monitoring in Digital Financial Systems

Read full text Summarise with VIRAT
Authors: Ashmitha Nagraj
Cited by: 0
References: 12

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

The rapid transformation of the world's financial system to digital payments and transactions has given rise to a corresponding increase in cyber-enabled financial crimes such as payment fraud, account takeover and synthetic identity theft. The traditional methods of detecting and preventing fraud rely on static rule-based systems, or linear statistical models which have been shown to be less effective at identifying sophisticated, non-linear fraud vectors and have high false positive rates. A Hybrid Real-Time Fraud Detection Framework (HRFDF), which integrates Long Short-Term Memory (LSTM) and Multi-Head Self-Attention Transformer Architectures is proposed for creating a Fraud Detection Engine with the ability to model Temporal Complexities in fraud patterns. To overcome the common challenges associated with real-time processing including latency and accuracy, the proposed system adopts a lambda architecture to process both streaming and batch data simultaneously. The two methods to solve this issue of imbalance classes are: Adaptive Synthetic Sampling (ADASYN) and focal loss function. Using the European credit card fraud dataset along with large amounts of synthetic data for transactions, the HRFDF was able to achieve fraud detection accuracy at levels of 99.92% and F1-score at levels of 96.8% [1]. In addition, SHapley Additive exPlanations (SHAP) is utilized to make sure that the models are explainable to the regulatory agencies.

Fields of Study & Subjects

Imbalanced Data Classification Techniques Privacy-Preserving Technologies in Data Cybercrime and Law Enforcement Studies

Publication details

Journal / source
European Modern Studies Journal
Volume
10
Issue
1
Pages
432–440
Publication date
23 March 2026
Language
EN