Credit Card Fraud Detection

Abstract

With the rise in financial transactions carried out digitally, crimes such as hacking, online frauds have also increased. Credit card fraud is one such crime. The project 'Credit Card Fraud Detection' is designed to overcome and minimize these kinds of frauds related to the transactions taking place via credit cards. Credit cards are used in shopping malls, gas stations, and also in online shopping or for paying the bills online. The project will assist in detecting fraudulent activities by distinguishing the normal transaction from the fraudulent transactions. Banks and Credit card companies currently detect fraud if the transaction is done from a remote place, or the amount is beyond the transactions carried out by the credit card holder in history. However, situations, when fraud is carried out from the jurisdiction of the customer and the amount, are also comparable to normal transactions of the customer are hard to detect. Thus, this project has utilized machine learning algorithms to predict fraud activities. The system is trained based on several different parameters in the transactions. We have utilized naive Bayes, Random Forest, and Adaptive boost algorithms to predict frauds with different accuracies. We have compared results from these algorithms on two different datasets. The results show different accuracies, giving us an insight that the size of data has an impact upon the percentile accuracy of all algorithms. However, it is important to point out that random forest performed better compared to other algorithms. The project is developed for the administrator at a financial institution or a bank where they can upload any dataset and perform prediction by utilizing the different algorithms. The User can also search a particular transaction through its id. The project contains an attractive user-friendly interface with these functionalities.