Title: Financial fraud identification technology based on FL model and MLP neural network

Authors: Zhenyu Chu; Xuewu Lai

Addresses: The Experimental Center, Liaoning University of International Business and Economics, Dalian, 116052, China ' The School of Business, Xi'an International University, Xi'an, 710077, China

Abstract: This study proposes a financial fraud identification system based on a federated learning (FL) model and a multi-layer perceptron (MLP) neural network. The system addresses data silos and class imbalance issues in financial institutions by using the Federated Average (FedAvg) algorithm for joint training and a composite minority class oversampling algorithm for data balancing. Experiments on a credit card transaction dataset show that the system achieves 99.60% training set accuracy and 99.21% test set accuracy, with a 2.15% average increase in area under the curve (AUC) and a 2.18% increase in geometric mean compared to other models. The results demonstrate the system's superior performance in financial fraud identification while ensuring user data privacy and security.

Keywords: federated machine learning; MLP; multi-layer perceptron; FFI; financial fraud identification; joint training; data balancing preprocessing.

DOI: 10.1504/IJTPM.2025.150705

International Journal of Technology, Policy and Management, 2025 Vol.25 No.4, pp.327 - 347

Received: 19 Sep 2024
Accepted: 22 Jan 2025
Published online: 22 Dec 2025 *

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