ONLINE CREDIT CARD FRAUD DETECTION USING GAUSSIAN MIXTURE MODELS

• Main Author: Low, Erick
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/39714
• Institution: Institut Teknologi Bandung

One of the most common credit card fraud types is called behavioral fraud. It happens when the detail information of one’s credit card has been obtained fraudulently, and transactions are made with that card. The cardholder is not aware at all when the transactions are made. This particular type of fraud easily happens in online transaction, like in e-commerce. It’s because, to make an online transaction, the frauder only requires the detail information of the credit card without has to own the card physically. This books attempts to provide a fraud detection method which can be used to check the transaction status, whether it’s legitimate or fraudulent. Normally, customers who are making transaction will have to pass transaction verification step. The transaction status will be checked by the fraud detection system. For customer service, we hope the verification step won’t be slow so that the customers won’t have to spend a long time in order to make a transaction. Given such situation, not only is a company required to provide a robust fraud detection system, but also a fast fraud detection system. To detect the fraudulent transactions, we’re interested in unsupervised method. We detect by observing the abnormal transactions that are being made. An abnormal transaction is transaction that doesn’t suit the customer’s past transaction behavior. This book uses Gaussian mixture models to do clustering. Clustering needs to be done in order for us to obtain several classes of customer transaction behavior. Later, we also give comparison between fraud detection method using Gaussian mixture models and k-means.