Fraud Detection in Credit Cards Using Methods of Machine Learning

Abstract

The rise of technological advancements and advanced communication networks leads to increase in fraud related to credit card. The repercussions of fraud related to credit card impacting both consumers and financial institutions. Fraudsters consistently evolve their techniques, emphasizing the necessity of making fraud protection technologies essential for banks and other financial entities. This research paper presents a method for an effective credit card fraud detection by integrating a feedback system using machine learning methodology. This feedback approach aims to enhance the detection accuracy and cost-effectiveness of the classifier. The study evaluates the performance of various methods, including artificial neural networks, random forest, Naive Bayes, tree classifiers, logistic regression, support vector machines, and gradient boosting classifiers. The evaluation is conducted on slightly skewed credit card fraud datasets containing transaction data from European account holders, totaling 284,807 trades. The evaluation considers both pre-processed content and raw. The efficiency of these methodologies is evaluated based on performance assessment dimensions for different classifiers, including precision, F1-score, accuracy, recall, and the false positive rate (FPR) percentage. The findings contribute to the ongoing efforts to develop robust systems for detecting and preventing fraud related to credit card, safeguarding from substantial financial harm.