Development of an Optimized Fraud Detection System Using Machine Learning in Financial Transactions
Abstract
Financial institutions together with their customers face substantial issues from evolving financial fraud because the annual worldwide financial loss amounts to hundreds of billions of dollars. The proposed research would implement an efficient system of detecting frauds using sophisticated machine learning methods, which would study the financial data to identify suspicious patterns. Our fraud detection system uses ensemble learning models with deep neural networks that unions supervised and unsupervised methods to solve the class imbalance challenges during fraud evaluation. Our proposed system delivers 97.2% accuracy alongside 94.1% precision while reaching 92.3% recall which surpasses both rule-based methodologies and independent machine learning solutions by 15–22%. The developed system brings vital advantages to financial institutions which help them cut fraudulent transaction losses while sustaining legitimate transaction capabilities for real-time transaction monitoring and emerging payment systems.