AI and ML Approaches for Credit Card Fraud Detection: A Comparative Study of Logistic Regression and Decision Tree Techniques

The article discusses the rising problem of fraud with credit cards, emphasizing how common it is becoming as online payment methods and e-commerce proliferate. Fraud detection has grown to be a significant concern due to the increase in online transactions, which has led researchers to investigate several machine learning methods. A unique fraud detection technique centered on transaction data is presented in this research. The strategy entails grouping cardholders according to the amount of money they spend, combining transactions within these groups using a window that moves technique, and examining historical transaction data to spot trends in customer behavior. Separate classifiers, including Decision Tree & Logistic Regression models, are trained using behavioral patterns that are taken from the aggregated data for each group. After assessing this and other models, the classification algorithm with the best performance is chosen for fraud prediction. The strategy also includes a feedback mechanism to handle idea drift, guaranteeing that the model adjusts to evolving patterns of transactions over time. The study's base is the European card fraud dataset, which offers a practical setting for validating the suggested approach.