-
J. J. Borges Oliveira
Universidade Federal Fluminense, Programa de Pós-graduação em Engenharia Elétrica e de Telecomunicações, Niterói, RJ
-
V. H. Ferreira
Universidade Federal Fluminense, Programa de Pós-graduação em Engenharia Elétrica e de Telecomunicações, Niterói, RJ
Keywords:
Supervised Learning, Classification Algorithms, Electricity distributors, System Distribution, Feature Extraction Methods, Identification, Fraud Reporting
Abstract
This article proposes the use of the supervised learning method to establish a scenario of classification and identification of possible customers of electric energy distributors with signs of irregularities. Using this method, the failures identified by energy utilities during the target generation process can be mapped, contributing to better assertiveness and decision making in the field. The model will use data belonging to a large electricity distributor in the state of Rio de Janeiro, to make a comparison between different processing and pre-processing approaches to point out evidence of fraud. For this, tests were carried out with different classifier algorithms and feature extraction methods, obtaining as a best result an average accuracy close to 70%.
