Title: A combined deep learning approach for fraudulent detection in the financial sector

Authors: Nripendra Narayan Das; C. Sivashanmugam; Sameer Shekhar

Addresses: Department of Information Technology, Manipal University Jaipur, Rajasthan, India ' Department of MBA, PES University, Bangalore, Karnataka, India ' School of Social, Financial & Human Sciences (KSFH), Kalinga Institute of Industrial Technology (KIIT) (Deemed to be University), Bhuvaneshvar, India

Abstract: One of the crucial topics in the banking sector is fraud detection (FD). The involvement of fraud has significantly increased as new technologies are introduced daily. The main reason why current algorithms cannot accurately identify fraud is because of ineffective feature learning and prediction methods. Most work only concentrated on a few parameters to identify fraud. In reality though, fraudsters quickly change their identities and other traits. By integrating the two data mining techniques of optimum feature learning (OFL) and precise classification methodologies, this research finds a solution to the issue. The tri-teaching learning (TTL) optimisation approach is used after the system initially gathers the financial data. Then, the fully recurrent neural network (FRNN) algorithm divides the data into legitimate and fraudulent categories. In trials, the proposed feature learning and deep detection methodology improves accuracy (around 98%), precision (93%), and recall (92%).

Keywords: fraud detection; data mining; deep learning; financial sector; feature extraction; fully recurrent neural network; FRNN; optimum feature learning; OFL; tri-teaching learning; TTL; quality education.

DOI: 10.1504/IJEF.2025.149159

International Journal of Electronic Finance, 2025 Vol.14 No.4, pp.502 - 523

Received: 19 May 2023
Accepted: 04 Sep 2023
Published online: 16 Oct 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article