Aplicação de tecnicas de mineração de dados para detectar fraudes na rede de distribuição de energia elétrica

Abstract

Computational intelligence is used through the application of a data mining model based on the process of knowledge discovery in databases, known as KDD. The model purpose is assist in solving one of the major problem faced by concessionaries of electric distribution throughout Brazil, which is the case of identifying fraudulent users on the electricity distribution networks. In this work, the objective is to identify user who do any type of fraud in the electricity distribution networks, seeking to reduce the losses due frauds and decrease the amount of spending time with inspections in the field. At the development of this research, were studied several techniques and algorithms of data mining, among them were chosen the algorithms of support vector machines and decision trees, which were managed through the WEKA data mining software, this software enables the training and validation of various data mining algorithms, to be used in the process of identifying user that must be inspected. This paper describes the data mining techniques that were used and also describes the case of study of an electric distribution company, where were applied the selection of users to be inspected due suspicion of fraud according to models of data mining used in this research. The results where satisfactory when evaluated with inspections in the field.