Análise discriminante aplicada na aprovação de pedidos realizados com cartão de crédito em um e-commerce

Abstract

The Brazilian e-commerce is growing fast and following this progress there is an increasing number of fraudulent practices that uses credit cards as form of payment to process online shopping. The aim of this study was to build a discriminative function that could be able to classify the order of payment method and so contribute with storeowners with cost reduction, fraud and the increase of customer confidence. The methodology applied was a Discriminant Analysis that is a technic that uses a Multivariate Analysis, where elements are classified from a sample or population in different groups. In this study, two distinct groups were used, that is, a low risk where we approve the order made by the customer and the high-risk, where the purchase is canceled. The data was supplied by a Brazilian e-commerce that gave a sample of 80 thousand company orders made in 2015. The discriminant function obtained in this study showed good results just to sort orders uncontested in the low risk group, not being effective in suspicious requests for classification in the high risk group.