Credit Card Fraud Detection using Autoencoders

Abstract

In today’s life or economy credit card plays a veryimportant role. Credit card becomes a necessary part of business, household and bank transactions. Using of credit card carefully and responsibly gives a enormous benefit to the user, fraudulent activities happen and give financial damage to the user or card holder. The growth of E-commerce industry led to use of credit card or many platform for online purchase or different transaction because of this the fraudulent activities was also increased. Bank facing many issues for detecting the credit card fraudulent transactions. For finding the fraud of credit card machine learning plays a vital role. For predicting these fraud detection of credit card we use many machine learning methods or algorithms, past data we collect and analyze it and make a machine learning model to detect the fraudulent activities which going to happen. The performance of fraud detecting in credit card transaction is greatly affected by sampling approach on data-set, selection of variable and detection technique used. This project objective to use of efficient approach to detect automatic fraud related to banks or insurance company using deep learning algorithm called autoencoder. We are using the European card holder real-time dataset of September 2013. The data was unbalanced in the dataset so for this autoencoder is perfect to provide the accurate results. We can reconstruct the normal data through the autoencoder and anomalies was detected at the time of reconstruction error threshold and consider the anomalies. Keyword: fraud detection, unsupervised learning, autoencoder, credit card