Comparative study of unsupervised and supervised learning analysis for the credit card fraud identification

Fraudulent activities are inherently unpredictable and inconsistent, making them challenging to detect. Their covert manipulation of authentication and security protocols incurs substantial financial losses, often amounting to millions of dollars. Credit card companies must be capable of identifying fraudulent transactions to prevent customers from being billed for unauthorized purchases. This study mainly aims in comparing the functioning of different machine learning models with credit card transaction dataset and relevant methodologies. Among the models evaluated, Random Forest achieved the highest accuracy, while Isolation Forest performed best within the group of unsupervised models. Overall, supervised models demonstrated superior performance compared to their unsupervised counterparts.