A Proposed Fraud Detection Model based on e-Payments Attributes a Case Study in Egyptian e-Payment Gateway

As per Payfort's 2017 report, titled State of payments in the Arab world; Egypt had a 22% yearly increase in the overall volume of internet payments in 2016, which was assessed at $6.2 billion. e-Payments are the major point of life nowadays in Egypt and the whole world; with tens of e-payments companies in Egypt and more than 5 million transactions done every day and 60 billion EGP volume of payments in 2018. Online and mobile fraud was estimated at $10.7 billion in 2015, as per Juniper Research, and is expected to reach $25.6 billion by the end of the decade. As the whole e-payments business is affected by fraud, e-payments firms and their consumers lose a lot of money. On the other hand, one of the most powerful techniques that could be used for fraud predictive is data mining techniques such as the decision tree. This paper introduces a prediction model for managing the risk of fraud in the Egyptian e-payment market that helps to reduce the loss of money. This model is developed using a real dataset from one of Egypt's top e-payment gateways based on the e-payment transaction attributes importance like transaction time, transaction amount, transaction limit, and transaction customer No. repetition limit. The importance of these attributes was determined using IBM SPSS modeler's decision tree and its predictors' importance. The model significantly assisted in the reduction of fraud cases by a very high rate, with an accuracy of 88.45% and a precision of 93.5% resulting in a savings of 101970.52 EGP out of 131297.83 EGP.