Fraud Detection in Credit Card Transactions using Machine Learning

Abstract

Credit card fraud has become very threatening to the financial market, with serious economic consequences. The developing nature of fraud activities is far beyond the capacity of simple traditional fraud-detection systems that depend on transaction volume or general strategies of fraudsters. As such, an improved fraud model that integrates machine learning classification technique and anomaly detection is proposed in this research. Our approach employs a combination of XGBoost, SMOTE, and Isolation Forest to handle class imbalance and improve the classification of fraudulent transactions. The model presented a relatively high value of precision(0.992), recall(0.1) and accuracy(0.1), and its validation did indicate competence in fraud detection that's accompanied by very minimal false positives.