Análisis de fraude en sistemas de pago electrónico en el sector empresarial de telecomunicaciones del Ecuador.

Abstract

The present study has been carried out to analyze fraud in electronic payment systems in the Ecuadorian telecommunications business sector, from an approach that promotes the improvement of trust in its customers and market segment. In this regard, it has been considered that most fraud detection systems (SDS) monitor the flows of credit card transactions through classifiers that return alerts for the riskiest payments. Fraud detection is remarkably a difficult problem due to the drift (CD) concept (that is, customer habits evolve) and class imbalance (that is, genuine transactions far outweigh frauds). In addition, the SDS differ from the conventional classification because, in the first phase, the researcher can only provide a small set of supervised samples that have time to evaluate only a small number of alerts. The methodology used is qualitative - quantitative, because the information provided in the study is qualified, through a review of the literature appropriate to the research and based on data from the organization (CLARO) to know the system fraud scenario Electronic payment Finally, the results were carried out with the design of two SDSs based on a set and a sliding window approach and it was shown that the winning strategy is to train two separate classifiers (in feedback and delayed labels, respectively), and then add the results.