Credit Card Fraudulent Transaction Detection Using Machine Learning Techniques

With the increasing use of credit cards in a cashless society, there has been a spike in fraudulent activities, resulting in severe losses for financial institutions. The need for more secure measures to restrict credit card fraud has become paramount to continue promoting the use of digital transactions. This paper provides an in-depth analysis of different techniques utilized to identify fraudulent transactions in credit card transactions. The aim is to differentiate between fraudulent and non-fraudulent activities. This paper suggested a new fraud detection approach for Streaming Transaction Data that analyzes customer transaction histories to identify behavioral patterns. The goal is to develop a novel method that improves current fraud prevention techniques. Various existing methods have been reviewed comprehensively to understand their efficacy. The paper explores different machine learning techniques such as DT, LR, SVM, XGBoost, and K nearest neighbor. A detailed analysis of the advantages and limitations of each method is provided, and the results are summarized. The suggested approach consists of two steps. Combining four of the top six machine learning algorithms with resampling techniques. In both stages, the Area under Accuracy and recall of each model are evaluated. The approach aims to enhance machine learning systems’ performance by incorporating resampling techniques. The paper concludes with recommendations on the best method to use depending on the specific application. The paper utilized the European credit card fraud dataset for its analysis and findings.