Credit Card Fraud – Comparative Study of the Effectiveness of Machine Learning Algorithms

Abstract

Credit cards offer convenience and user-friendly options for everyday transactions, attracting a wide audience seeking hassle-free financial interactions. The continuous evolution of technology and expansive network coverage facilitates easier access to credit cards and encourages their frequent usage among citizens. However, alongside the growth of the credit card industry, the threat of fraud looms large. One prevalent form of fraud involves unauthorized use of credit card details for purchases without the card owner's consent. With the sheer volume of card transactions, financial institutions and credit card issuers face significant challenges in detecting these fraudulent activities. This research aims to investigate the potential of Machine Learning techniques in identifying fraudulent transactions within a given dataset of credit card transactions. Additionally, the study assesses various Machine Learning algorithms using a publicly available credit card transactions dataset, analyzing their effectiveness in distinguishing between fraudulent and genuine transactions. Machine Learning algorithms like “Logistic Regression”, “Gaussian Naïve Bayes”, “KNeighbors Classifier”, “Linear Support Vector Classifier”, “Random Forest Classifier”, “Isolation Forest”, “Bagging Classifier”, “Decision Tree Classifier”, “Keras Classifier”, “MLP Classifier”, “LightGBM Classifier”, “XGboost Classifier”, “Adaboost Classifier”, “Catboost Classifier”, “Dynamic Ensemble” and “Stacking” are applied on the available dataset and compared for various performance metrics.