Fraud-Auditor: A Machine Learning Based System for Financial Fraud Detection

With the rapid increase in digital transactions, fi- nancial fraud has become a serious concern for organizations and individuals. Traditional rule-based systems fail to detect complex and evolving fraud patterns efficiently. This paper presents Fraud-Auditor, a machine learning-based system designed to detect fraudulent transactions using multiple classification al- gorithms such as Support Vector Machine (SVM), Multi-Layer Perceptron (MLP), Logistic Regression, and Gradient Boosting. The system incorporates data preprocessing, feature engineering, and real-time prediction through a web-based interface. Experi- mental results demonstrate an accuracy of up to 95.22%, making the system reliable for fraud detection. Additionally, evaluation techniques such as confusion matrix and ROC curve are used to analyze performance in detail.