Credit Card Fraud Identification using Logistic Regression and Random Forest

Fraud is an ancient yet ever-changing profession. Because of the digitization of money, financial transactions, banks, fraudsters now have a limitless number of possibilities to perpetrate crime from behind a screen, anywhere around the world. Fraud has a broad influence, with direct ramifications for business and the economy. It is of great worry to cybercrime organizations as recent studies have proven that ML algorithms may successfully be utilized to identify fraudulent transactions in massive amounts of payment data. Such techniques may identify fraudulent transactions in real time, which human auditors may miss. In this research, we apply supervised ML algorithms to the issue of fraud identification by analyzing simulated financial transaction data that is available to the public. Our aim is to show how supervised ML methods may be utilized to successfully identify data with extreme class disproportion. By way of example, we show how exploratory analysis may be utilized to identify fraudulent from real purchases. We also show that Random Forest outperform Logistic Regression when applied to a clearly distinguished dataset.