Machine learning methods for credit card fraud detection

Abstract

The rising popularity of digital banking has reshaped the financial environment and at the same time exposed the institutions to escalating cybersecurity threats. With the growth of online transactions, so does the difficulty of identifying and stopping fraudulent acts and this is where the system on traditional fails to identify the dynamic and intricate patterns of fraud in real time which makes it need smarter and adaptable solutions. In this study, the capabilities of machine learning models were explored to enhance cybersecurity on an anonymized European credit card fraud dataset on Kaggle, a collection of machine learning algorithms as well as Decision Tree, Logistic Regression, XGBoost, Random Forest, LightGBM, CatBoost, and Multi-Layer Perceptron were implemented and tested. The most useful model used in detecting fraud was identified based on key performance indicators like F1-score, precision, recall, and ROC-AUC. Overall, the research indicated that properly trained ML models may distinguish between legitimate and fraud transactions with a high level of accuracy. Overall, the study shows the significance of machine learning as an efficient instrument of improving cyber resilience in the age of digital finance.