EVALUATION OF THE EFFECTIVENESS OF MACHINE LEARNING TECHNIQUES FOR DETECTING CREDIT CARD FRAUD

Abstract

Every year, fraudulent credit card transactions result in losses of billions of dollars.
In order to minimise these losses, it is crucial to develop effective fraud detection algorithms. The
unpredictable distribution of the data, the extremely imbalanced class distributions, and the scarcity
of transactions with fraud investigators' labels make the construction of fraud detection algorithms
particularly difficult.In today's financial industry, credit card fraud is a steadily rising issue.
Fraudulent actions have increased quickly in frequency during the past few years, costing many
businesses, organisations, and government bodies a sizable amount of money.As a result of the
predicted growth in numbers, several academics in this area have concentrated on applying cuttingedge
machine learning approaches to identify fraudulent behaviours early on.It's simple and easy to
fall victim to credit card theft. Researchers began utilising various machine learning techniques to
detect and analyse scams in online transactions as fraud rates increased. .The groups are then used
to train various classifiers individually in the future. The classifier with the highest rating score can
then be selected as one of the most effective ways to detect fraud. Consequently, a feedback system
is used to address the issue of notion drift. In this study, we used data on European credit card theft.