Diseño de un modelo predictivo a través de la técnica de minera de datos ‘Random Forest’ para la detección de fraude bypass en redes telefónicas en el Ecuador.

Abstract

Telecommunications companies use greater configurations and elements in order to keep their growing demand of keep us communicated. These sometimes generates gaps that are used by organizations or people to carry out illegal actions such as bypass fraud. It fact, 5% of the mobile services companies’s revenues in the country are lost due to the lack of detection of such events. Which causes revenue leaks for organizations that offer this service worldwide and nationally. This work called “Design of a predictive model through data mining technique 'Random Forest' for the detection of bypass fraud in telephone networks in Ecuador” aims to build a predictive model using data mining through of the KDD methodology (“Discovery of Knowledge in Database”) in order to contributes to the effectiveness in the detection of this type of fraud. The methodology is applied using the KNIME software tool by implementing a workflow with the use of random forests as a supervised classification data mining technique where call records are used as basis for transforming them into a suitable mining view for the construction of the model. The results of the work indicated that data mining use reported greater efficiency than a traditional CDRs analysis in the detection of cases of bypass fraud and sets up the basis for future studies of the topic in this business.