Generating Personalized Algorithms to Learn Bayesian Network Classifiers for Fraud Detection in Web Transactions

Abstract

The volume of electronic transactions has raised a lot in last years, mainly due to the popularization of e-commerce. We also observe a significant increase in the number of fraud cases, resulting in billions of dollars losses each year worldwide. Therefore, it is essential to develop and apply techniques that can assist in fraud detection. In this direction, we propose an evolutionary algorithm to automatically build Bayesian Network Classifiers (BNCs) tailored to solve the problem of detecting fraudulent transactions. BNCs are powerful classification models that can deal well with data features, missing data and uncertainty. In order to evaluate the techniques, we adopt an economic efficiency metric and apply them to our real dataset. Our results show good performance in fraud detection, presenting gains up to 17%, compared to the actual scenario of the company.