A Comparative Scrutiny on Machine Learning and Deep Learning Approaches for Fraudulent Transaction Discovery in Credit Card Data

Over the last few years, use of credit card for online purchase has increased exponentially, which resulted in increase in fraud related to it. It is very difficult to detect fraudulent transactions from the banking transactions. This paper presents a comparative analysis between machine and deep learning techniques for detecting fraudulent transactions in credit card data. To develop a better method for detecting fraud, we use actual “credit card” data and apply a DL model called Long Short-Term Memory (LSTM). The efficiency of this approach is gauged with logistic regression and eXtreme gradient boosting algorithms of machine learning. Experimental evaluation reveals that the proposed LSTM-based approach achieves the excellent outcome in terms of ROC-AUC (100%), recall (100%), F1-score (99.97%), accuracy (99.97%), and precision (99.95%), surpassing the other methods. By utilizing the strengths of LSTM, which is particularly skilled at recognizing patterns and relationships over prolonged intervals in sequential data, the proposed approach effectively detects fraudulent transactions. This research demonstrates how DL techniques could improve fraud detection systems and provides valuable insights for creating effective models and improving security measures in real-time financial transactions. Study makes a great impact to the field of credit card fraudulent transaction detection by offering a comparative scrutiny of ML techniques with DL technique.