Performance Analysis of Credit Card Fraud Analysis and Detection Machine Learning Algorithms

Credit card fraud has become increasingly common despite advancements in credit card technology aimed at enhancing security. Various forms of fraud, such as stolen or lost cards, as well as hacking into digital payment methods like Apple Pay and Google Pay, contribute to this issue. As a result, it is crucial to detect and prevent credit card fraud, particularly from the perspective of credit card issuers. Machine Learning offers a promising approach to address this challenge, specifically through the application of classification algorithms. These algorithms can analyze a set of variables or a single variable and assign it to a particular class, making them suitable for identifying fraudulent and non-fraudulent credit card transactions. This research paper presents the outcomes of implementing multiple machine learning algorithms to classify credit card transactions as either fraudulent or non-fraudulent. The dataset used in this study comprises credit card transactions made by European cardholders during a two-day period in September 2013. We start with analysing the data of the transactions, building the classification models for detecting fraud, comparing the results of the models built, comparing the performance of the best model with the models researched previously and finally choosing the best model for the fraud detection. In this research paper, we used various classification algorithms and to improve the accuracy of the models, we had used three class balancing techniques which are Random Oversampling, SMOTE and ADASYN. We have found that Random Forest model with ADASYN class imbalancing technique is the best model for fraud detection as it has and ROC-AUC score of 0.99 on the test dataset.