Cost Reduce: Credit Card Fraud Identification Using Machine Learning

Nowadays, one of the most serious ethical challenges in the financial arena is payment card skimming. Payment card fraud comes in many forms. It is a major and long-term menace to society, with a significant economic effect. With the advancement of contemporary technology, including worldwide interaction, fraud is on the rise. As a result, combating fraud has become a critical topic to investigate. There are a variety of machine learning methods that can assist us in classifying unusual or fraudulent transactions. The only requirements are historical data and an appropriate algorithm that can better match our data. The main agenda of the papers is to improve costs by reducing baseline with a new sample process, discuss the different choices of the optimization metric, and try to show the best performing models for fraud detection. For more details, this research study has used Baseline logistic regression, Gradient boosted model (GbM), Logistic regression, Logistic Regression Baseline Model (LgBm), Extreme Gradient (XGBoost) classifier with Synthetic Minority Oversampling Technique (SmOtE) and Adaptive Synthetic Sampling Method (AdAsYn), showed the accuracy. In the end, this research study reminds you that fraud, which represents 0.1731% of the studied dataset. The findings and conclusions are based on real-world transactional data given by a big European card processing business.

I. Introduction

Payment cards are indeed a quick and easy way to make transactions. Fraud cases are on the rise as a result of the increased use of payment cards, particularly in online purchases [2] [8]. The growth generates financial risk and uncertainty, as its outcomes in billions of dollars/euros in losses each year in the commercial sector. However, real transaction records that might aid in the creation of efficient fraud detection prediction models are difficult to come by, owing to concerns about consumer information confidentiality.