Credit Card Fraud Detection Using Autoencoder Algorithm with SMOTE Technique

In today’s life credit cards play a significant role in many people’s lives. The credit card market has exploded over the years, thanks in large part to the computerization of society. In fact, credit cards are the payment method with the highest reported instances of fraud. Some criminals commit fraud by using lost or stolen credit cards. Others make illegal transactions without even having the credit card in their pocket they can access the victim’s funds with just a basic knowledge of the card or account credentials. This project intends to detect credit card fraud. So, our objective is to first solve the problem of the unbalanced data by applying the SMOTE (Synthetic Minority Over-sampling Technique) technique and then using a deep learning algorithm called autoencoder. In this study we used a public dataset collected in September 2013 by European cardholders with 284,807 transactions, of which 492 were fraudulent.