Financial fraud identification and economic impact analysis based on machine vision technology

Abstract

Financial fraud is one of the primary threats to businesses and the economy at large, resulting in monumental financial losses, operational inefficiency, and significant damage to reputation. This research examines the use of a Multi-Technique Fraud Detection System (MTFDS) that incorporates machine vision technologies such as Optical Character Recognition (OCR), signature verification, and behavioral anomaly detection to identify and prevent financial fraud and analyze its economic impact. The fraud detection dataset contains detailed transaction data from scanned financial documents. Noise reduction is applied to the scanned financial documents using a median filter for improved OCR processing. OCR extracts and validates text from scanned invoices and bank statements. Signature verification ensures the authenticity of signed documents. A Modified White Shark Optimizer tuned Elman Spiking Neural Network (MWSO-ESNN) is proposed to detect behavioral anomalies by identifying irregular transaction patterns and visual anomalies in transaction-related images. The research also assesses the economic consequences of fraud by quantifying financial losses, operational costs, reputation damage, and market instability. Experimental results indicate that the MWSO-ESNN model achieves 99.2% accuracy, 96.7% precision, 92.5% recall, and 95.6% F1-score, outperforming Deep Neural Networks (DNNs) and Long Short-Term Memory (LSTM) networks. This research demonstrates the efficacy of MTFDS in preventing financial fraud and provides a comprehensive assessment of its economic effects, offering a holistic approach to fraud detection and economic impact analysis.